Deploy optimized Arcade Hub with shared models and LFS binaries

Concurrency limits, session timeout, shared MediaPipe models, static
frontend split, health/metrics endpoints, and Git LFS for image/apk assets.

.gitattributes

@@ -38,3 +38,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.png filter=lfs diff=lfs merge=lfs -text
 *.wav filter=lfs diff=lfs merge=lfs -text
 *.mp3 filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+*.avif filter=lfs diff=lfs merge=lfs -text
+*.apk filter=lfs diff=lfs merge=lfs -text
+*.JPEG filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -1,6 +1,6 @@
 FROM python:3.11-slim
 
-# Install system dependencies required by OpenCV headless, Pygame, and MediaPipe
+# System deps for OpenCV headless, Pygame, MediaPipe + CJK fonts
 RUN apt-get update && apt-get install -y --no-install-recommends \
     libglib2.0-0 \
     libgl1 \
@@ -8,31 +8,33 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     libgles2 \
     libegl1 \
     fonts-wqy-zenhei \
-    fontconfig \
+    fonts-wqy-microhei \
+    fonts-noto-cjk \
     && rm -rf /var/lib/apt/lists/*
 
-# Set up environment variables
 ENV PYTHONUNBUFFERED=1 \
     PYTHONDONTWRITEBYTECODE=1 \
     SDL_VIDEODRIVER=dummy \
     SDL_AUDIODRIVER=dummy \
-    PORT=7860
+    PORT=7860 \
+    MAX_CONCURRENT_RUNNERS=3 \
+    SESSION_TIMEOUT_SEC=900 \
+    PROJECT_CACHE_TTL_SEC=10
 
-# Create working directory and copy dependencies
 WORKDIR /app
 COPY requirements.txt .
 
-# Install dependencies
+# Base image without TensorFlow (smaller / faster cold start).
+# To include Keras demos: COPY requirements-ml.txt . && pip install -r requirements-ml.txt
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Copy application files
 COPY . .
 
-# Create directory for projects if it doesn't exist
-RUN mkdir -p /app/projects
+RUN mkdir -p /app/projects /app/models /app/static \
+    && chmod -R 777 /app
 
-# Set permissions for Hugging Face Spaces compatibility (runs as UID 1000)
-RUN chmod -R 777 /app
+# Healthcheck for orchestrators
+HEALTHCHECK --interval=30s --timeout=5s --start-period=40s --retries=3 \
+    CMD python -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:%s/health' % __import__('os').environ.get('PORT','7860'))" || exit 1
 
-# Run server.py on port specified by environment
 CMD uvicorn server:app --host 0.0.0.0 --port $PORT

models/README.md

@@ -0,0 +1,20 @@
+# Shared MediaPipe models
+
+Large `.task` model files live here once and are linked into student project
+directories at runtime by `runner.py` when missing.
+
+| File | Used by |
+|------|---------|
+| `hand_landmarker.task` | Hand / pinch / gesture projects |
+| `face_landmarker.task` | Face mesh / eye-distance projects |
+
+Student code may still use:
+
+```python
+os.path.join(os.path.dirname(__file__), "hand_landmarker.task")
+```
+
+The runner creates a symlink (or hard link) from the project folder to this
+directory before executing `main.py`.
+
+Do not commit duplicate copies under `projects/**`.

projects/2026_AI營1/劉宗穎/DataFrame指令练习/d1_1_pandas_intro.py

@@ -0,0 +1,37 @@
+# d1_1_pandas_intro.py
+# Day 1 學員填空：Pandas 基礎（完成 TODO 後執行）
+# cd examples/Day1/student && python3 d1_1_pandas_intro.py
+# 參考答案：../Teacher/d1_1_pandas_intro.py
+
+import pandas as pd
+
+df = pd.read_csv("taipei_weather_2026.csv")
+print("=== 數據載入成功！ ===")
+
+print("\n--- 1. head ---")
+print(df.head())
+
+print("\n--- 2. info ---")
+df.info()
+
+print("\n--- 3. describe ---")
+print(df.describe())
+
+print("\n--- 4. Temperature 前 5 筆 ---")
+print(df['Temperature'].head())
+
+print("\n--- 5. Date 與 Rainfall ---")
+print(df[['Rainfall', 'Date']].head())
+
+print("\n--- 挑戰：最高溫與平均濕度 ---")
+max_temp = df['Temperature'].max()
+avg_humidity = df['Humidity'].mean()
+min_rainfall = df['Rainfall'].min()
+print(f"最高溫度：{max_temp}°C")
+print(f"平均濕度：{avg_humidity:.2f}%")
+print(f"最低降雨:{min_rainfall:.2f}%")
+
+print("\n--- whether it is good for egyptians")
+
+egy = df[(df['Temperature'] > 27) & (df['Rainfall'] == 0)] 
+print(egy)

projects/2026_AI營1/劉宗穎/DataFrame指令练习/taipei_weather_2026.csv

@@ -0,0 +1,16 @@
+Date,Temperature,Humidity,Rainfall
+2026-07-01,28.5,82,5.2
+2026-07-02,29.0,80,0.0
+2026-07-03,30.2,75,0.0
+2026-07-04,31.5,70,0.0
+2026-07-05,27.8,90,25.4
+2026-07-06,26.5,95,42.0
+2026-07-07,28.0,85,12.5
+2026-07-08,29.5,78,0.0
+2026-07-09,30.8,72,0.0
+2026-07-10,32.0,68,0.0
+2026-07-11,28.2,88,18.0
+2026-07-12,27.0,92,30.5
+2026-07-13,29.2,80,2.1
+2026-07-14,30.0,76,0.0
+2026-07-15,31.2,71,0.0

projects/2026_AI營1/劉宗穎/Day1/taipei_weather_2026.csv

@@ -0,0 +1,16 @@
+Date,Temperature,Humidity,Rainfall
+2026-07-01,28.5,82,5.2
+2026-07-02,29.0,80,0.0
+2026-07-03,30.2,75,0.0
+2026-07-04,31.5,70,0.0
+2026-07-05,27.8,90,25.4
+2026-07-06,26.5,95,42.0
+2026-07-07,28.0,85,12.5
+2026-07-08,29.5,78,0.0
+2026-07-09,30.8,72,0.0
+2026-07-10,32.0,68,0.0
+2026-07-11,28.2,88,18.0
+2026-07-12,27.0,92,30.5
+2026-07-13,29.2,80,2.1
+2026-07-14,30.0,76,0.0
+2026-07-15,31.2,71,0.0

projects/2026_AI營1/劉宗穎/ai_api串接练习/d2_1_gemini.py

@@ -0,0 +1,41 @@
+# d2_1_gemini.py — Day2 學員填空
+# cd examples/Day2/Student && python3 d2_1_gemini.py
+# 參考答案：../Teacher/d2_1_gemini.py
+
+import os
+from google import genai
+
+# === 【學員填空練習點 1：設置與載入 API Key】 ===
+# TODO: 嘗試從環境變數 GEMINI_API_KEY 取得 API Key。如果未設定，則使用 input() 提示手動輸入。
+api_key = os.environ.get("GEMINI_API_KEY", "AQ.Ab8RN6LgriztRWiwQOdkK1wXmmfs4AzlZKSQsP2X0cnjzaMCCg")
+
+if not api_key:
+    print("[提示] 未檢測到環境變數 GEMINI_API_KEY。")
+    print("      請先設定：export GEMINI_API_KEY=\"your_key\"")
+    print("      或於下方手動輸入（僅供本機測試，勿寫死在檔案中上傳）：")
+    api_key = input("API Key: ").strip()
+    if not api_key:
+        print("[結束] 未提供 API Key。可改練習 d2_1_ollama_local.py。")
+        exit(0)
+
+# === 【學員填空練習點 2：設定模型名稱】 ===
+# TODO: 設定使用的 Gemini 模型名稱（例如 'gemini-3.1-flash-lite'）
+model_name = "gemini-3.1-flash-lite"
+
+print(f"正在連線至 Google Cloud Model ({model_name})...")
+
+# === 【學員填空練習點 3：初始化 Gemini Client】 ===
+# TODO: 呼叫 genai.Client 並傳入 api_key 參數來初始化 Client 物件，指派給 client 變數
+client = genai.Client(api_key=api_key)
+
+# === 【學員填空練習點 4：發送內容生成請求】 ===
+# TODO: 使用 client.models.generate_content 呼叫模型，傳入 model 參數與 contents 提示詞
+#       提示詞內容：請模型以繁體中文解釋什麼是『API 串接』，並用餐廳點餐作為比喻，字數約 100 字。
+response = client.models.generate_content(model=model_name, contents="请用中文解释为什么泡面很好吃，字数一百字以内")
+# === 【學員填空練習點 5：讀取並印出回應】 ===
+# TODO: 從 response 物件中取得 text 欄位內容，並列印出來。
+print("\n=== AI 回應結果 ===")
+# 請在下方實作列印邏輯：
+print(response.text)
+# for i in range(10):
+#     print(response.text)

projects/2026_AI營1/劉宗穎/mock&AI_api双模式/d2_3_chatbot_memory.py

@@ -0,0 +1,95 @@
+# d2_3_chatbot_memory.py
+# Day 2: 生成式 AI 串接實戰 - 具對話記憶的智慧客服機器人 (Gemini / Local Mock 雙模)
+
+import os
+from google import genai
+
+# 優先讀取環境變數 GEMINI_MODEL，預設為 gemini-3.1-flash-lite
+GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-3.1-flash-lite")
+api_key = os.environ.get(
+    "GEMINI_API_KEY", "")
+
+
+def local_mock_response(user_input, history):
+    user_input_lower = user_input.lower()
+
+    # TODO: 處理 營業時間 / 你好 / 再見，其餘回兜底句
+    if "營業時間" in user_input_lower or "時間" in user_input_lower:
+        return "我們觀測站的營業時間為週一至週五的上午 9:00 至下午 6:00，週末休息。"
+    # elif "XX" in user_input_lower or "XX" in user_input_lower:
+    #     return "XXXXXXXXXXXXXXXXX"
+
+
+    else:
+        turns = len(history) // 2 + 1
+        return (
+            f"（模擬模式）收到您的訊息：『{user_input}』。"
+            f"這是我與您的第 {turns} 次對話，請問還有其他需要協助的嗎？"
+        )
+
+
+def main():
+    print("=" * 50)
+    print("  得能 AI 智慧客服對話系統 (支援對話上下文記憶)")
+    print("=" * 50)
+
+    use_mock = False
+    chat_session = None
+
+    if not api_key:
+        print("[提示] 未檢測到環境變數 GEMINI_API_KEY，將自動啟用「離線模擬模式 (Local Mock Mode)」。")
+        print("      （如需連線雲端模型，請設定 GEMINI_API_KEY 環境變數）")
+        use_mock = True
+    else:
+        try:
+            print(f"正在連線 Google Gemini 雲端客服伺服器 ({GEMINI_MODEL})...")
+            client = genai.Client(api_key=api_key)
+            # TODO: 設定系統指令 (System Instruction / System Prompt)
+            # system_instruction = (
+            #     "你是一位專業的得能教育客服人員，語氣溫和客氣，"
+            #     "XXXXXXXXXXXXXXXXXXXXXXX"
+            # )
+
+            chat_session = client.chats.create(
+                model=GEMINI_MODEL,
+                config={'system_instruction': system_instruction}
+            )
+            print("雲端客服系統已就緒！")
+        except Exception as e:
+            print(f"連線雲端客服失敗 ({e})，切換至離線模擬模式...")
+            use_mock = True
+
+    local_history = []
+
+    print("\n系統啟動完成！請輸入您要諮詢的問題（輸入 'quit' 或 'exit' 結束對話）：")
+    print("-" * 50)
+
+    while True:
+        try:
+            user_input = input("\n[您]: ").strip()
+            if not user_input:
+                continue
+
+            if user_input.lower() in ["quit", "exit"]:
+                print("對話結束，感謝您的使用！")
+                break
+
+            if use_mock:
+                ai_reply = local_mock_response(user_input, local_history)
+                local_history.append({"role": "user", "text": user_input})
+                local_history.append({"role": "model", "text": ai_reply})
+            else:
+                response = chat_session.send_message(user_input)
+                ai_reply = response.text
+
+            print(f"[AI 客服]: {ai_reply}")
+
+        except KeyboardInterrupt:
+            print("\n對話強行終止。再見！")
+            break
+        except Exception as e:
+            print(f"\n[錯誤] 發生異常：{e}")
+
+
+if __name__ == "__main__":
+    main()

projects/2026_AI營1/劉宗穎/mock&AI_api双模式/d2_4_chatbot_memory.py

@@ -0,0 +1,129 @@
+# d2_3_chatbot_memory.py
+# Day 2: 生成式 AI 串接實戰 - 具對話記憶的智慧客服機器人 (Gemini / Local Mock 雙模)
+
+import os
+from google import genai
+
+# 優先讀取環境變數 GEMINI_MODEL，預設為 gemini-3.1-flash-lite
+GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-3.1-flash-lite")
+api_key = os.environ.get(
+    "GEMINI_API_KEY", "AQ.Ab8RN6JvbUqLW-wcddeMducmta0aGE2QtcrmF4Kfd8nnpZ9mzg")
+
+
+def local_mock_response(user_input, history):
+    user_input_lower = user_input.lower()
+
+    if "營業時間" in user_input_lower or "時間" in user_input_lower:
+        return "我们冰淇淋店的營業時間為週一至週五的上午 9:00 至下午 6:00，週末休息。"
+    elif "地址" in user_input_lower or "地点" in user_input_lower or "地方" in user_input_lower:
+         return "我们的店在求毛星球毛笔城市QQ镇喜喜街167号一楼。"
+    elif "冰淇淋的口味" in user_input_lower or "口味" in user_input_lower:
+        return "我们的冰淇淋的口味有草莓,芒果,巧克力,咖啡,蓝莓,汽水和芭乐。"
+    elif "冰淇淋的价格" in user_input_lower or "价格" in user_input_lower or "冰淇淋怎么卖" in user_input_lower:
+        return "一球100元,两球170元,三球250元,五球400元,十球750元。"
+    elif "有打折吗" in user_input_lower or "有优惠吗" in user_input_lower or "有进行促销活动吗" in user_input_lower or "有活动吗" in user_input_lower or "折扣" in user_input_lower:
+        return "我们有和吃鸡炸鸡店合作，去那里消费满1000元即可换一球冰淇淋。"
+    elif "help" in user_input_lower or "帮助" in user_input_lower:
+        return "\n輸入「營業時間、時間」即可知道營業時間\n輸入「地址、地點、地方」即可知道地址\n輸入「冰淇淋的口味、口味」即可知道口味\n輸入「冰淇淋的價格、價格、冰淇淋怎麼賣」即可知道價格\n輸入「有打折嗎、有優惠嗎、有進行促銷活動嗎、有活動嗎、折扣」即可知道優惠資訊\n輸入「停車場、停車、附近有沒有停車場」即可知道停車資訊\n輸入「可以客製化嗎、客製化的服務、客製化」即可知道客製化服務\n輸入「客製化的價錢、客製化要多少錢」即可知道客製化價格\n輸入「推薦、招牌」即可知道推薦口味\n輸入「外帶、內用」即可知道用餐方式\n輸入「付款、支付、可以刷卡」即可知道付款方式。"
+    elif "你好" in user_input_lower or "hello" in user_input_lower or "hi" in user_input_lower or "Hi" in user_input_lower:
+        return "你好！请问您需要什么帮助吗？"
+    elif "停车场" in user_input_lower or "停车" in user_input_lower or "附近有没有停车场" in user_input_lower:
+        return "我们的店附近有三个停车场，步行三分钟即可到达我们的店。"
+    elif "可以克制化吗" in user_input_lower or "客制化的服务"  in user_input_lower or "客制化"  in user_input_lower:
+        return "我们的店可以克制化冰淇淋,可以选择彩糖、珍珠、草莓酱、巧克力酱、布丁、奶油和炼乳。"
+    elif "客制化的价钱" in user_input_lower or "客制化要多少钱" in user_input_lower or "客制化要多少钱？" in user_input_lower:
+        return "客制化每一种材料都要再加15元。"
+    elif "推薦" in user_input_lower or "招牌" in user_input_lower:
+        return "推薦您試試草莓、巧克力和芭樂口味，都是熱門選擇。"
+    elif "外帶" in user_input_lower or "內用" in user_input_lower:
+        return "我們提供內用及外帶服務。"
+    elif "付款" in user_input_lower or "支付" in user_input_lower or "可以刷卡" in user_input_lower:
+        return "我們接受現金、信用卡及行動支付。"
+    elif "谢谢" in user_input_lower or "Thank you" in user_input_lower:
+        return "不客气！还需要什么帮助吗?"
+
+    else:
+        # 計算目前是第幾輪對話（一輪包含使用者輸入與客服回覆，共 2 個元素）
+        # Why: 藉由傳入的 history 串列長度除以 2，即可動態得知目前互動的輪數，模擬記憶狀態。
+        turns = len(history) // 2 + 1
+        return (
+            f"（模擬模式）收到您的訊息：『{user_input}』。"
+            f"這是我與您的第 {turns} 次對話，請問還有其他需要協助的嗎？"
+        )
+
+
+def main():
+    print("=" * 50)
+    print("   吃吃吃冰淇淋店客服對話系統 (支援對話上下文記憶)")
+    print("=" * 50)
+
+    use_mock = False
+    chat_session = None
+
+    if not api_key:
+        print("[提示] 未檢測到環境變數 GEMINI_API_KEY，將自動啟用「離線模擬模式 (Local Mock Mode)」。")
+        print("      （如需連線雲端模型，請設定 GEMINI_API_KEY 環境變數）")
+        use_mock = True
+    else:
+        try:
+            print(f"正在連線 Google Gemini 雲端客服伺服器 ({GEMINI_MODEL})...")
+            client = genai.Client(api_key=api_key)
+            # TODO: 設定系統指令 (System Instruction / System Prompt)
+            system_instruction = (
+                "你是一位冰淇淋店的客服，负责提供资讯"
+                "我们冰淇淋店的營業時間為週一至週五的上午 9:00 至下午 6:00，週末休息。"
+                "我们的店在求毛星球毛笔城市QQ镇喜喜街167号一楼。"
+                "冰淇淋的价格一球100元,两球170元,三球250元,五球400元,十球750元。"
+                "我们的店有进行优惠，有和吃鸡炸鸡店合作，去那里消费满1000元即可换一球冰淇淋。"
+                "这是我们的资讯：輸入「營業時間、時間」即可知道營業時間,輸入「地址、地點、地方」即可知道地址,輸入「冰淇淋的口味、口味」即可知道口味,輸入「冰淇淋的價格、價格、冰淇淋怎麼賣」即可知道價格,輸入「有打折嗎、有優惠嗎、有進行促銷活動嗎、有活動嗎、折扣」即可知道優惠資訊,輸入「停車場、停車、附近有沒有停車場」即可知道停車資訊,輸入「可以客製化嗎、客製化的服務、客製化」即可知道客製化服務,輸入「客製化的價錢、客製化要多少錢」即可知道客製化價格,輸入「推薦、招牌」即可知道推薦口味,輸入「外帶、內用」即可知道用餐方式,輸入「付款、支付、可以刷卡」即可知道付款方式。"
+                "你应该亲切的向客户问好。"
+                "我们的店附近有三个停车场，步行三分钟即可到达我们的店。"
+                "我们的店可以克制化冰淇淋,可以选择彩糖、珍珠、草莓酱、巧克力酱、布丁、奶油和炼乳。"
+                "客制化每一种材料都要再加15元。"
+                "我們提供內用及外帶服務。"
+                "我們接受現金、信用卡及行動支付。"
+            )   
+
+            chat_session = client.chats.create(
+                model=GEMINI_MODEL,
+                config={'system_instruction': system_instruction}
+            )
+            print("雲端客服系統已就緒！")
+        except Exception as e:
+            print(f"連線雲端客服失敗 ({e})，切換至離線模擬模式...")
+            use_mock = True
+
+    local_history = []
+
+    print("\n系統啟動完成！請輸入您要諮詢的問題（輸入 'quit' 或 'exit' 結束對話）：")
+    print("-" * 50)
+
+    while True:
+        try:
+            user_input = input("\n[您]: ").strip()
+            if not user_input:
+                continue
+
+            if user_input.lower() in ["quit", "exit"]:
+                print("對話結束，感謝您的使用！")
+                break
+
+            if use_mock:
+                ai_reply = local_mock_response(user_input, local_history)
+                local_history.append({"role": "user", "text": user_input})
+                local_history.append({"role": "model", "text": ai_reply})
+            else:
+                response = chat_session.send_message(user_input)
+                ai_reply = response.text
+
+            print(f"[AI 客服]: {ai_reply}")
+
+        except KeyboardInterrupt:
+            print("\n對話強行終止。再見！")
+            break
+        except Exception as e:
+            print(f"\n[錯誤] 發生異常：{e}")
+
+
+if __name__ == "__main__":
+    main()

projects/2026_AI營1/劉宗穎/mock_关键词抓取/d2_3_chatbot_memory_mock.py

@@ -0,0 +1,84 @@
+# d2_3_chatbot_memory_mock.py
+# Day 2: 生成式 AI 串接實戰 - 具對話記憶的智慧客服機器人 (純 Local Mock 教師版)
+
+def local_mock_response(user_input: str, history: list) -> str:
+    """
+  _  根據使用者輸入與歷史對話紀錄，模擬客服助理的自動回覆。
+    在沒有網路或 API 金鑰時，如何透過關鍵字比對
+    與對話歷史長度（計算對話輪數）來模擬一個具有「記憶」能力的對話流程。
+    """
+    user_input_lower = user_input.lower()
+
+    # 關鍵字比對：模擬常見的客服問答
+    # TODO: 處理 營業時間 / 你好 / 再見，其餘回兜底句
+    if "營業時間" in user_input_lower or "時間" in user_input_lower:
+        return "我们冰淇淋店的營業時間為週一至週五的上午 9:00 至下午 6:00，週末休息。"
+    elif "地址" in user_input_lower or "地点" in user_input_lower or "地方" in user_input_lower:
+         return "我们的店在求毛星球毛笔城市QQ镇喜喜街167号一楼。"
+    elif "冰淇淋的口味" in user_input_lower or "口味" in user_input_lower:
+        return "我们的冰淇淋的口味有草莓,芒果,巧克力,咖啡,蓝莓,汽水和芭乐。"
+    elif "冰淇淋的价格" in user_input_lower or "价格" in user_input_lower or "冰淇淋怎么卖" in user_input_lower:
+        return "一球100元,两球170元,三球250元,五球400元,十球750元。"
+    elif "有打折吗" in user_input_lower or "有优惠吗" in user_input_lower or "有进行促销活动吗" in user_input_lower or "有活动吗" in user_input_lower or "折扣" in user_input_lower:
+        return "我们有和吃鸡炸鸡店合作，去那里消费满1000元即可换一球冰淇淋。"
+    elif "help" in user_input_lower or "帮助" in user_input_lower:
+        return "\n輸入「營業時間、時間」即可知道營業時間\n輸入「地址、地點、地方」即可知道地址\n輸入「冰淇淋的口味、口味」即可知道口味\n輸入「價格、冰淇淋怎麼卖、冰淇淋的价格」即可知道價格\n輸入「折扣、優惠、活動」即可知道優惠資訊。"
+    elif "你好" in user_input_lower or "hello" in user_input_lower or "hi" in user_input_lower or "Hi" in user_input_lower:
+        return "你好！请问您需要什么帮助吗？"
+        
+    else:
+        # 計算目前是第幾輪對話（一輪包含使用者輸入與客服回覆，共 2 個元素）
+        # Why: 藉由傳入的 history 串列長度除以 2，即可動態得知目前互動的輪數，模擬記憶狀態。
+        turns = len(history) // 2 + 1
+        return (
+            f"（模擬模式）收到您的訊息：『{user_input}』。"
+            f"這是我與您的第 {turns} 次對話，請問還有其他需要協助的嗎？"
+        )
+
+
+def main():
+    print("=" * 50)
+    print("  吃吃吃冰淇淋店客服對話系統 (純離線模擬模式 - 支援對話記憶)")
+    print("=" * 50)
+    print("[提示] 目前為純 Local Mock 模式，不需設定環境變數與連線雲端 API。")
+    print("-" * 50)
+
+    # 用於儲存對話歷史的串列
+    local_history = []
+
+    print("\n系統啟動完成！請輸入您要諮詢的問題（輸入 'quit' 或 'exit' 結束對話）：")
+    print("-" * 50)
+    print("如果你不知道要输入什么，输入help或者帮助。")
+    print("-" * 50)
+
+    while True:
+        try:
+            user_input = input("\n[您]: ").strip()
+            if not user_input:
+                continue
+
+            # 結束條件判斷
+            if user_input.lower() in ["quit", "exit"]:
+                print("對話結束，感謝您的使用！")
+                break
+
+            # 呼叫模擬回覆函式
+            ai_reply = local_mock_response(user_input, local_history)
+
+            # 將本次對話記錄寫入歷史紀錄中，維持記憶狀態
+            # Why: 每次成功的問答都必須 append 進 history，以便下一輪對話時能正確累計輪數。
+            local_history.append({"role": "user", "text": user_input})
+            local_history.append({"role": "model", "text": ai_reply})
+
+            print(f"[AI 客服]: {ai_reply}")
+
+        except KeyboardInterrupt:
+            # 處理使用者按下 Ctrl+C 強行結束的情境，避免程式崩潰
+            print("\n對話強行終止。再見！")
+            break
+        except Exception as e:
+            print(f"\n[錯誤] 發生異常：{e}")
+
+
+if __name__ == "__main__":
+    main()

projects/2026_AI營1/劉宗穎/mock_密室逃脱游戏/d2_3_chatbot_memory_game_template.py

@@ -0,0 +1,118 @@
+# d2_3_chatbot_memory_game_template.py
+# Day 2: 生成式 AI 串接實戰 - 離線 Mock 遊戲設計模板 (學生填空版)
+
+def local_mock_response(user_input: str, history: list) -> str:
+    """
+    根據使用者輸入與歷史對話紀錄，模擬 NPC 或是系統的對話反應。
+    
+    Why:
+    此檔案為困難版對話記憶解謎的設計模板。你將學習到：
+    1. 前置狀態限制：必須先問過「線索一」，才能查到「線索二」...
+    2. 動態 Flag：Flag 答案中的某個線索會隨提問次數（turns）改變。
+    """
+    user_input_lower = user_input.strip().lower()
+    
+    # 計算目前是第幾輪對話
+    turns = len(history) // 2 + 1
+    
+    # TODO: 設計你的動態變數（例如 turns * 5，或是 current_weather 等）
+    # 並設定最終預期的正確 Flag 答案（必須包含此動態變數）
+    expected_flag = "flag}.{...add_8_sprinkles_ice_cream_Chocolate_flavored...}.{,"
+
+    # 從歷史中判斷先前是否已取得前置線索（在 history 中，AI model 是否曾回覆過特定字眼）
+    # TODO: 依據你設計的線索關鍵字，修改以下 history 的 any 比對條件
+    asked_clue1 = any("Chocolate_flavored" in h["text"] for h in history if h["role"] == "model")
+    asked_clue2 = any("_ice_cream_" in h["text"] for h in history if h["role"] == "model")
+
+    # 1. 驗證 Flag 邏輯
+    if user_input_lower == expected_flag:
+        return f"🎉 恭喜你！成功破解案件並獲得了神秘的冰淇淋配方！你在第 {turns} 回合成功通關！"
+    
+    elif "flag{" in user_input_lower:
+        return (
+            f"❌ Flag 驗證失敗！\n"
+            f"提示：最後一個線索會隨著提問次數動態改變。\n"
+            f"你現在是在第 {turns} 次提問輸入，因此當前的動態線索應該是 turns{turns} 喔！\n"
+            f"請重新計算並拼湊最新的 Flag！"
+        )
+
+    # 2. 劇情關鍵字問答分支（打造你的 NPC 對話樹！）
+    # TODO: 依據你的故事背景，修改並設計以下 if-elif-else 結構
+    if "flag" in user_input_lower or "答案" in user_input_lower:
+        return "NPC：『想要解開這起謎題，你必須向我打聽案子的「原因」、「重要物品」與「凶器」！』"
+
+    elif "原因" in user_input_lower or "谁做的" in user_input_lower:
+        # 線索一：靜態線索
+        return "【線索一】\n这个案件是暖暖火锅店做的，\n因为他说每次顾客都说吃完了火锅很冷，\n可是他们吃完就去吃冰淇淋，所以才会感觉很冷。\n有一天，他就去了吃吃吃冰淇淋店复仇，\n让他们的老板踩到掉到地上的巧克力冰淇淋，并把他绑起来了。\nChocolate_flavored,记住这个口味！"
+
+    elif "重要物品" in user_input_lower or "物品" in user_input_lower or "重要的物品" in user_input_lower:
+        # 線索二：有前置限制，必須先取得線索一
+        if not asked_clue1:
+            return "NPC：『你連发生了什么事都不知道，我才不告訴你重要物品是什么呢！』"
+        return "【線索二】\n他在绑架了老板以后，他就把神秘的冰淇淋配方藏了起来。\n_ice_cream_, 记住这个线索！"
+
+    elif "凶器" in user_input_lower or "武器" in user_input_lower or "線索三" in user_input_lower:
+        # 線索三：動態線索，包含 turns 變數，且必須先取得線索二
+        if not asked_clue1:
+            return "NPC：『你連发生了什么事都不知道，我才不告訴你凶器是什么呢！』"
+        if not asked_clue2:
+            return "NPC：『你必須先查出「重要的物品」，我才會告訴你凶器是什么。』"
+        return f"【線索三】犯人用的凶器是一种配料，在里面加了安眠药，所以最后一个线索是add_8_sprinkles。"
+    elif "hint" in user_input_lower:
+        return "321"
+
+    elif "拼湊" in user_input_lower or "結合" in user_input_lower or "怎麼組合" in user_input_lower:
+        return "提示：將【線索一】【線索二】【當下提問的線索三】依序結合，並用小寫的 flag}.{... ...}.{,包起來。"
+
+    # 3. 兜底回覆（當玩家輸入沒有匹配到任何關鍵字時）
+    else:
+        return (
+            f"NPC：『你在說什麼？我聽不懂。收到您的訊息：『{user_input}』。"
+            f"這是我與您的第 {turns} 次對話，請問還有其他需要協助的嗎？』"
+        )
+
+    
+
+def main():
+    # TODO: 修改為你的遊戲標題與故事背景
+    print("=" * 60)
+    print("      🕵️  我設計的第一個具解謎遊戲 - 冰淇淋店店长失踪案  🕵️")
+    print("=" * 60)
+    print("本次密室逃脱体验与吃吃吃冰淇淋店合作")
+    print("=" * 60)
+    print("[故事簡介] 你是一位偵探，必須利用歷史對話記憶拼湊線索。")
+    print("傳說中，有一家聞名世界的冰淇淋店，憑藉一份神秘配方製作出無人能複���的美味冰淇淋。\n然而，某天店長突然失蹤，珍貴的冰淇淋秘方也隨之消失，只留下一間被上鎖的店鋪和一連串神秘線索。\n你們是一群受邀前來協助調查的人，當踏入店內的瞬間，大門突然自動鎖上，並響起一段錄音：\n「只有找到真正的秘方繼承者，才能打開大門離開這裡。」\n店內藏有許多與冰淇淋、口味、價格、優惠活動及客製化相關的謎題。\n你們必須仔細觀察每個角落、破解一道道機關，蒐集所有線索，\n找到失落的冰淇淋秘方，才能成功逃離這間神秘的冰淇淋店。")
+    print("依序追查主角、地點與動態蹤跡。祝你逃脱成功！")
+    print("-" * 60)
+
+    local_history = []
+
+    print("\n遊戲已啟動！請輸入你想問 NPC 的話（輸入 'quit' 結束遊戲）：")
+    print("-" * 60)
+
+    while True:
+        try:
+            user_input = input("\n[您]: ").strip()
+            if not user_input:
+                continue
+
+            if user_input.lower() in ["quit", "exit"]:
+                print("遊戲結束，感謝遊玩！")
+                break
+
+            ai_reply = local_mock_response(user_input, local_history)
+
+            local_history.append({"role": "user", "text": user_input})
+            local_history.append({"role": "model", "text": ai_reply})
+
+            print(f"[NPC]: {ai_reply}")
+
+        except KeyboardInterrupt:
+            print("\n遊戲強行終止！")
+            break
+        except Exception as e:
+            print(f"\n[錯誤] 發生異常：{e}")
+
+
+if __name__ == "__main__":
+    main()

projects/2026_AI營1/劉宗穎/requirements.txt

@@ -0,0 +1,6 @@
+pandas>=2.0.0
+matplotlib>=3.7.0
+requests>=2.31.0
+beautifulsoup4>=4.12.0
+opencv-python>=4.8.0
+google-generativeai>=0.8.0

projects/2026_AI營1/劉宗穎/三指捏游戏/d5_pinch_challenge_solution.py

@@ -0,0 +1,217 @@
+"""
+Week 10 - 魔法手勢：Pinch 偵測與手勢按鈕 (三指捏合挑戰 Template 填空版)
+==================================================================
+本週挑戰目標：
+  1. 實作「三指捏合 (Three-Finger Pinch)」：
+     必須同時滿足「大拇指 (4) 與食指 (8) 的距離 < Threshold」
+     以及「大拇指 (4) 與中指 (12) 的距離 < Threshold」才判定為 Pinch
+  2. 用三指碰觸 Pygame 畫面上的按鈕 + 捏合 = 魔法點擊
+  3. 使用邊緣觸發 (Edge Detection) 防止重複觸發
+  4. 新增 30 秒限時機制，時間到顯示最終得分
+"""
+import pygame
+import cv2
+import mediapipe as mp
+import math
+import sys
+import os
+import random
+
+# ============================================
+# 1. 初始化
+# ============================================
+pygame.init()
+WIDTH, HEIGHT = 800, 600
+screen = pygame.display.set_mode((WIDTH, HEIGHT))
+pygame.display.set_caption("Week 10 - 三指魔法手勢按鈕 (填空版) ")
+clock = pygame.time.Clock()
+from mediapipe.tasks import python
+from mediapipe.tasks.python import vision
+
+# 攝影機
+cap = cv2.VideoCapture(0)
+
+# 【TODO 1】初始化 MediaPipe 手勢偵測器 (Hand Landmarker)
+\
+
+try:
+    from mediapipe.tasks import python
+    from mediapipe.tasks.python import vision
+    base_options = python.BaseOptions(model_asset_path=os.path.join('hand_landmarker.task'))
+    options = vision.HandLandmarkerOptions(base_options=base_options, num_hands=1)
+    detector = vision.HandLandmarker.create_from_options(options)
+except Exception as e:
+    print(f"Error loading model: {e}")
+    pygame.quit()
+    sys.exit()
+
+# ============================================
+# 2. 遊戲變數
+# ============================================
+# 手指座標
+finger_x, finger_y = 400, 300   # 食指
+thumb_x, thumb_y = 400, 300     # 大拇指
+middle_x, middle_y = 400, 300   # 中指
+
+# 距離變數
+dist_thumb_index = 0.0
+dist_thumb_middle = 0.0
+
+# Pinch 相關變數
+is_pinching = False
+was_pinching = False     # 邊緣觸發
+PINCH_THRESHOLD = 50     # 距離臨界值
+
+# 按鈕設定
+button_rect = pygame.Rect(300, 250, 200, 100)
+button_colors = [(100, 100, 220), (220, 100, 100),
+                 (100, 220, 100), (220, 180, 50)]
+current_color_index = 0
+
+# 點擊計數器
+click_count = 0
+
+# --- 限時機制變數 ---
+START_TIME = pygame.time.get_ticks()
+GAME_DURATION = 30  # 遊戲限制時間（秒）
+game_state = "PLAYING"  # 狀態機
+
+# ============================================
+# 3. 主迴圈
+# ============================================
+running = True
+while running:
+    for event in pygame.event.get():
+        if event.type == pygame.QUIT:
+            running = False
+
+    # --- 讀取攝影機與手勢偵測 ---
+    success, img = cap.read()
+    is_pinching = False  # 每幀重設狀態，防止手離屏時卡死
+
+    if success:
+        img = cv2.flip(img, 1)  # 鏡像翻轉
+        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+
+        mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+        results = detector.detect(mp_image)
+
+        # === 【TODO 2：挑戰關卡三之手部安全偵測】 ===
+        # 1. 判斷手部是否在畫面中 (results.hand_landmarks 是否有資料)
+        # 2. 如果在畫面中，取得食指 (8)、大拇指 (4)、中指 (12) 的像素座標
+        # 3. 如果手不在畫面中，將 is_pinching 設為 False
+        #
+        hand_in_frame = False
+        if results.hand_landmarks:
+            hand_in_frame = True
+            for hand_lms in results.hand_landmarks:
+                # 食指 (8)
+                lm8 = hand_lms[8]
+                finger_x = int(lm8.x * WIDTH)
+                finger_y = int(lm8.y * HEIGHT)
+
+                # 大拇指 (4)
+                lm4 = hand_lms[4]
+                thumb_x = int(lm4.x * WIDTH)
+                thumb_y = int(lm4.y * HEIGHT)
+
+                # 中指 (12)
+                lm12 = hand_lms[12]
+                middle_x = int(lm12.x * WIDTH)
+                middle_y = int(lm12.y * HEIGHT)
+
+                # === 【TODO 3：挑戰關卡一之三指捏合判定】 ===
+                # 提示：計算食指與大拇指距離，以及中指與大拇指距離
+                # 兩組距離皆小於 PINCH_THRESHOLD 時，is_pinching 才為 True
+                dist_thumb_index = math.hypot(finger_x - thumb_x, finger_y - thumb_y)
+                dist_thumb_middle = math.hypot(middle_x - thumb_x, middle_y - thumb_y)
+                is_pinching = (dist_thumb_index < PINCH_THRESHOLD) and (dist_thumb_middle < PINCH_THRESHOLD)
+
+                dist_thumb_index = math.hypot(finger_x - thumb_x, finger_y - thumb_y)
+                dist_thumb_middle = math.hypot(middle_x - thumb_x, middle_y - thumb_y)
+                is_pinching = (dist_thumb_index < PINCH_THRESHOLD) and (dist_thumb_middle < PINCH_THRESHOLD)
+
+    # --- 狀態機邏輯處理 ---
+    if game_state == "PLAYING":
+        # 1. 計算剩餘時間
+        current_ticks = pygame.time.get_ticks()
+        elapsed_seconds = (current_ticks - START_TIME) / 1000
+        remaining_time = max(0, GAME_DURATION - elapsed_seconds)
+
+        # 2. 檢查時間是否到了
+        if remaining_time <= 0:
+            game_state = "GAME_OVER"
+
+        # 3. 邊緣觸發點擊判斷
+        if is_pinching and not was_pinching:
+            finger_rect = pygame.Rect(finger_x - 5, finger_y - 5, 10, 10)
+            if finger_rect.colliderect(button_rect):
+                click_count += 1
+                current_color_index = (current_color_index + 1) % len(button_colors)
+                
+                # === 【TODO 4：挑戰關卡二之按鈕隨機移動】 ===
+                # 提示：當點擊發生時，讓按鈕的 x 和 y 座標隨機分配在畫面內
+
+                button_rect.x = random.randint(0, WIDTH - button_rect.width)
+                button_rect.y = random.randint(0, HEIGHT - button_rect.height)
+                print(f" 魔法三指點擊！第 {click_count} 次")
+
+        # 更新上一幀狀態
+        was_pinching = is_pinching
+
+    # --- 繪製畫面 ---
+    screen.fill((20, 20, 35))
+    font = pygame.font.SysFont(None, 32)
+    large_font = pygame.font.SysFont(None, 64)
+
+    if game_state == "PLAYING":
+        # 繪製按鈕
+        btn_color = button_colors[current_color_index]
+        pygame.draw.rect(screen, btn_color, button_rect, border_radius=16)
+
+        btn_text = font.render(f"Clicks: {click_count}", True, (255, 255, 255))
+        screen.blit(btn_text, (button_rect.centerx - btn_text.get_width() // 2,
+                               button_rect.centery - btn_text.get_height() // 2))
+
+        # 顯示倒數計時
+        time_text = font.render(f"Time Left: {remaining_time:.1f}s", True, (255, 220, 100))
+        screen.blit(time_text, (WIDTH - time_text.get_width() - 20, 10))
+
+        # 畫手指游標
+        cursor_color = (50, 255, 50) if is_pinching else (255, 50, 50)
+        pygame.draw.circle(screen, cursor_color, (finger_x, finger_y), 12)    # 食指
+        pygame.draw.circle(screen, (255, 255, 255), (thumb_x, thumb_y), 12)  # 大拇指
+        pygame.draw.circle(screen, (100, 150, 255), (middle_x, middle_y), 12) # 中指
+
+        # === 【TODO 5：挑戰關卡三之畫面提示】 ===
+        # 提示：當手不在畫面中時，在畫面上方顯示警告文字提醒學生將手部放回
+       
+        if not hand_in_frame:
+            warning_text = font.render("OUT OF FRAME - PLEASE PUT HANDS BACK", True, (255, 100, 100))
+            screen.blit(warning_text, (WIDTH // 2 - warning_text.get_width() // 2, 100))
+
+        # 顯示調試數值
+        small_font = pygame.font.SysFont(None, 24)
+        status = small_font.render(f"Finger: ({finger_x}, {finger_y})", True, (150, 150, 150))
+        screen.blit(status, (10, 10))
+
+        dist_text = small_font.render(f"Dist Thumb-Index: {dist_thumb_index:.0f} px", True, (150, 150, 200))
+        screen.blit(dist_text, (10, 35))
+
+        dist_text2 = small_font.render(f"Dist Thumb-Middle: {dist_thumb_middle:.0f} px", True, (150, 150, 200))
+        screen.blit(dist_text2, (10, 60))
+
+    elif game_state == "GAME_OVER":
+        # 繪製遊戲結束畫面
+        over_text = large_font.render("GAME OVER", True, (255, 50, 50))
+        score_text = font.render(f"Total Clicks: {click_count} times", True, (255, 255, 255))
+
+        screen.blit(over_text, (WIDTH // 2 - over_text.get_width() // 2, HEIGHT // 2 - 50))
+        screen.blit(score_text, (WIDTH // 2 - score_text.get_width() // 2, HEIGHT // 2 + 20))
+
+    pygame.display.update()
+    clock.tick(30)
+
+cap.release()
+pygame.quit()
+sys.exit()

projects/2026_AI營1/劉宗穎/人脸辨识/d3_2_face_detection.py

@@ -0,0 +1,128 @@
+# d3_2_face_detection.py
+# Day 3: 電腦視覺與影像辨識 - Haar Cascade 即時人臉偵測與相機串流 (學員練習版)
+import cv2
+import os
+import urllib.request
+
+# 1. 取得 Haar Cascade 臉部特徵分類器模型檔案
+XML_FILE = "haarcascade_frontalface_default.xml"
+XML_URL = (
+    "https://raw.githubusercontent.com/opencv/opencv/master/"
+    "data/haarcascades/haarcascade_frontalface_default.xml"
+)
+
+# 優先使用 OpenCV 內建路徑，其次本機檔案，最後才下載
+
+
+def load_face_cascade():
+    bundled = os.path.join(cv2.data.haarcascades,
+                           "haarcascade_frontalface_default.xml")
+    if os.path.exists(bundled):
+        print(f"使用 OpenCV 內建分類器：{bundled}")
+        return cv2.CascadeClassifier(bundled)
+
+    if not os.path.exists(XML_FILE):
+        print("未檢測到臉部特徵分類模型檔，正在從 OpenCV 官方來源下載...")
+        try:
+            urllib.request.urlretrieve(XML_URL, XML_FILE)
+            print("分類器下載完成！")
+        except Exception as e:
+            print(f"[錯誤] 下載模型檔失敗：{e}")
+            print("請手動下載 haarcascade_frontalface_default.xml 並置於同目錄下。")
+            exit(1)
+
+    cascade = cv2.CascadeClassifier(XML_FILE)
+    if cascade.empty():
+        print("[錯誤] 分類器載入失敗，XML 可能損毀。")
+        exit(1)
+    return cascade
+
+
+face_cascade = load_face_cascade()
+
+# 2. 啟動視訊鏡頭 (0 代表預設鏡頭)
+# 無鏡頭時可改為影片：cv2.VideoCapture("test_video.mp4")
+print("正在開啟視訊鏡頭，請稍候...")
+cap = cv2.VideoCapture(0)
+
+if not cap.isOpened():
+    print("[錯誤] 無法開啟視訊鏡頭！請確認鏡頭權限已開啟且未被其他程式佔用。")
+    print("備案：將 VideoCapture(0) 改為 VideoCapture(\"test_video.mp4\") 使用影片測試。")
+    exit(1)
+
+print("\n視訊串流已啟動！")
+print("提示：點擊視窗後，按下鍵盤上的 'q' 鍵即可退出程式。")
+print("-" * 50)
+
+while True:
+    ret, frame = cap.read()
+    if not ret:
+        print("[警告] 無法讀取鏡頭畫面，正在重新嘗試...")
+        continue
+
+    # === 【重要觀念 TODO 1：水平翻轉鏡頭畫面】 ===
+    # 觀念重點：預設讀取的鏡頭畫面是左右相反的（像看著別人）。
+    # 使用 cv2.flip(frame, 1) 可以將畫面水平翻轉，使其符合我們照鏡子的視覺習慣。
+    # 請解開下方註釋 (刪除行首的 #) 並執行，觀察鏡頭畫面的翻轉變化。
+    frame = cv2.flip(frame, 1)
+
+    # 預設使用原始畫面，避免未解開註釋時程式出錯
+    gray = frame
+
+    # === 【重要觀念 TODO 2：灰階轉換 (色彩空間轉換)】 ===
+    # 觀念重點：人臉特徵偵測 (如 Haar Cascade) 主要依賴亮度與對比資訊，不需要色彩資訊。
+    # 將影像轉換為灰階 (BGR to GRAY) 可以減少 2/3 的資料量，大幅提升偵測速度並減少干擾。
+    # 請解開下方註釋 (刪除行首的 #)，將畫面轉換為灰階。
+    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+
+    # 預設無偵測到人臉，避免未解開註釋時程式出錯
+    faces = []
+
+    # === 【重要觀念 TODO 3：使用 Haar Cascade 分類器進行人臉偵測】 ===
+    # 觀念重點：detectMultiScale 會在影像中尋找臉部，回傳每張臉的 (x, y, w, h) 座標。
+    # - scaleFactor=1.1: 每次搜尋時將影像縮小 10%，用來偵測不同大小（距離）的人臉。
+    # - minNeighbors=5: 判定同一區域為人臉的鄰近候選框數量，數值愈高愈嚴格，能減少誤判。
+    # - minSize=(60, 60): 忽略小於此尺寸的偵測框，避免遠處背景雜訊干擾。
+    # 請解開下方所有註釋 (刪除行首的 #) 來啟用人臉偵測功能。
+    faces = face_cascade.detectMultiScale(
+        gray,
+        scaleFactor=1.1,
+        minNeighbors=1,
+        minSize=(60, 60),
+    )
+
+    # === 【重要觀念 TODO 4：繪製人臉偵測外框與文字標記】 ===
+    # 觀念重點：當模型找出人臉 (x, y, w, h) 之後，我們可以在畫面上把結果畫出来。
+    # - cv2.rectangle: 畫矩形，需要指定左上角 (x, y) 與右下角 (x+w, y+h)、BGR 顏色與框線粗細。
+    # - cv2.putText: 寫文字，常用於分類結果或標記名稱。
+    # 請解開下方所有註釋 (刪除行首的 #)，將偵測到的結果畫在畫面上。
+    for (x, y, w, h) in faces:
+        cv2.circle(frame, (x + w // 2, y + h // 2), 15, (0, 0, 255), -1)
+        cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 100, 0), 3)
+        cv2.putText(
+            frame, "a big idiot", (x, y - 10),
+            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (255, 100, 0), 2,
+        )
+
+    # 畫面左上角顯示目前狀態：有抓到臉就顯示數量，否則顯示搜尋中
+    if len(faces) > 0:
+        cv2.putText(
+            frame, f"Faces Detected: {len(faces)}", (20, 40),
+            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 2,
+        )
+    else:
+        cv2.putText(
+            frame, "Searching Face...", (20, 40),
+            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2,
+        )
+
+    # 把本回合處理後的影格送到視窗，形成即時串流效果
+    cv2.imshow("Smart Face Detection Camera", frame)
+
+    # waitKey(1) 既檢測按鍵，也驅動視窗重繪
+    if cv2.waitKey(1) & 0xFF == ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()
+print("視訊鏡頭已關閉，資源已安全釋放。")

projects/2026_AI營1/劉宗穎/人脸辨识/d3_2_face_detection2.py

@@ -0,0 +1,139 @@
+# d3_2_face_detection.py
+# Day 3: 電腦視覺與影像辨識 - Haar Cascade 即時人臉偵測與相機串流 (學員練習版)
+import cv2
+import os
+import urllib.request
+
+# 1. 取得 Haar Cascade 臉部特徵分類器模型檔案
+XML_FILE = "haarcascade_frontalface_default.xml"
+XML_URL = (
+    "https://raw.githubusercontent.com/opencv/opencv/master/"
+    "data/haarcascades/haarcascade_frontalface_default.xml"
+)
+
+# 優先使用 OpenCV 內建路徑，其次本機檔案，最後才下載
+
+
+def load_face_cascade():
+    bundled = os.path.join(cv2.data.haarcascades,
+                           "haarcascade_frontalface_default.xml")
+    if os.path.exists(bundled):
+        print(f"使用 OpenCV 內建分類器：{bundled}")
+        return cv2.CascadeClassifier(bundled)
+
+    if not os.path.exists(XML_FILE):
+        print("未檢測到臉部特徵分類模型檔，正在從 OpenCV 官方來源下載...")
+        try:
+            urllib.request.urlretrieve(XML_URL, XML_FILE)
+            print("分類器下載完成！")
+        except Exception as e:
+            print(f"[錯誤] 下載模型檔失敗：{e}")
+            print("請手動下載 haarcascade_frontalface_default.xml 並置於同目錄下。")
+            exit(1)
+
+    cascade = cv2.CascadeClassifier(XML_FILE)
+    if cascade.empty():
+        print("[錯誤] 分類器載入失敗，XML 可能損毀。")
+        exit(1)
+    return cascade
+
+
+face_cascade = load_face_cascade()
+
+# 2. 啟動視訊鏡頭 (0 代表預設鏡頭)
+# 無鏡頭時可改為影片：cv2.VideoCapture("test_video.mp4")
+print("正在開啟視訊鏡頭，請稍候...")
+cap = cv2.VideoCapture(0)
+
+if not cap.isOpened():
+    print("[錯誤] 無法開啟視訊鏡頭！請確認鏡頭權限已開啟且未被其他程式佔用。")
+    print("備案：將 VideoCapture(0) 改為 VideoCapture(\"test_video.mp4\") 使用影片測試。")
+    exit(1)
+
+print("\n視訊串流已啟動！")
+print("提示：點擊視窗後，按下鍵盤上的 'q' 鍵即可退出程式。")
+print("-" * 50)
+
+while True:
+    ret, frame = cap.read()
+    if not ret:
+        print("[警告] 無法讀取鏡頭畫面，正在重新嘗試...")
+        continue
+
+    # === 【重要觀念 TODO 1：水平翻轉鏡頭畫面】 ===
+    # 觀念重點：預設讀取的鏡頭畫面是左右相反的（像看著別人）。
+    # 使用 cv2.flip(frame, 1) 可以將畫面水平翻轉，使其符合我們照鏡子的視覺習慣。
+    # 請解開下方註釋 (刪除行首的 #) 並執行，觀察鏡頭畫面的翻轉變化。
+    frame = cv2.flip(frame, 1)
+
+    # 預設使用原始畫面，避免未解開註釋時程式出錯
+    gray = frame
+
+    # === 【重要觀念 TODO 2：灰階轉換 (色彩空間轉換)】 ===
+    # 觀念重點：人臉特徵偵測 (如 Haar Cascade) 主要依賴亮度與對比資訊，不需要色彩資訊。
+    # 將影像轉換為灰階 (BGR to GRAY) 可以減少 2/3 的資料量，大幅提升偵測速度並減少干擾。
+    # 請解開下方註釋 (刪除行首的 #)，將畫面轉換為灰階。
+    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+
+    # 預設無偵測到人臉，避免未解開註釋時程式出錯
+    faces = []
+
+    # === 【重要觀念 TODO 3：使用 Haar Cascade 分類器進行人臉偵測】 ===
+    # 觀念重點：detectMultiScale 會在影像中尋找臉部，回傳每張臉的 (x, y, w, h) 座標。
+    # - scaleFactor=1.1: 每次搜尋時將影像縮小 10%，用來偵測不同大小（距離）的人臉。
+    # - minNeighbors=5: 判定同一區域為人臉的鄰近候選框數量，數值愈高愈嚴格，能減少誤判。
+    # - minSize=(60, 60): 忽略小於此尺寸的偵測框，避免遠處背景雜訊干擾。
+    # 請解開下方所有註釋 (刪除行首的 #) 來啟用人臉偵測功能。
+    faces = face_cascade.detectMultiScale(
+        gray,
+        scaleFactor=1.1,
+        minNeighbors=5,
+        minSize=(60, 60),
+    )
+
+    # === 【重要觀念 TODO 4：繪製人臉偵測外框與文字標記】 ===
+    # 觀念重點：當模型找出人臉 (x, y, w, h) 之後，我們可以在畫面上把結果畫出来。
+    # - cv2.rectangle: 畫矩形，需要指定左上角 (x, y) 與右下角 (x+w, y+h)、BGR 顏色與框線粗細。
+    # - cv2.putText: 寫文字，常用於分類結果或標記名稱。
+    # 請解開下方所有註釋 (刪除行首的 #)，將偵測到的結果畫在畫面上。
+    for (x, y, w, h) in faces:
+        #1. 擷取人臉區域 (Slice)
+        face_roi = frame[y:y+h, x:x+w]
+
+        #2. 縮小人臉圖像 (例如縮小到 15x15 像素
+        small = cv2.resize(face_roi, (15, 15), interpolation=cv2.INTER_LINEAR)
+
+        #3. 放大回原本的大小 (使用 NEAREST 內插法保留顆粒感)
+        mosaic = cv2.resize(small, (w, h), interpolation=cv2.INTER_NEAREST)
+
+        #4. 放回原畫面中
+        frame[y:y+h, x:x+w] = mosaic
+
+        cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 100, 0), 3)
+        cv2.putText(
+            frame, "dumboo dumboo", (x, y - 10),
+            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (255, 100, 0), 2,
+        )
+
+    # 畫面左上角顯示目前狀態：有抓到臉就顯示數量，否則顯示搜尋中
+    if len(faces) > 0:
+        cv2.putText(
+            frame, f"Faces Detected: {len(faces)}", (20, 40),
+            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 2,
+        )
+    else:
+        cv2.putText(
+            frame, "Searching Face...", (20, 40),
+            cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2,
+        )
+
+    # 把本回合處理後的影格送到視窗，形成即時串流效果
+    cv2.imshow("Smart Face Detection Camera", frame)
+
+    # waitKey(1) 既檢測按鍵，也驅動視窗重繪
+    if cv2.waitKey(1) & 0xFF == ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()
+print("視訊鏡頭已關閉，資源已安全釋放。")

projects/2026_AI營1/劉宗穎/咖啡店分析图表/coffee_sales_data.csv

@@ -0,0 +1,15 @@
+Date,Sales_Amount,Customer_Count,Coffee_Cups_Sold,Dessert_Sold,Promotion_Active
+2026-07-01,15200,120,85,30,No
+2026-07-02,16800,135,95,35,No
+2026-07-03,24500,190,140,55,Yes
+2026-07-04,28000,210,165,68,Yes
+2026-07-05,18500,145,110,40,No
+2026-07-06,14200,110,80,25,No
+2026-07-07,15900,125,90,28,No
+2026-07-08,17200,138,100,32,No
+2026-07-09,26000,205,150,60,Yes
+2026-07-10,29500,225,175,75,Yes
+2026-07-11,21000,160,120,45,No
+2026-07-12,19800,150,115,42,No
+2026-07-13,15000,118,85,26,No
+2026-07-14,16500,130,92,30,No

projects/2026_AI營1/劉宗穎/咖啡店分析图表/d1_4_visualization_coffee.py

@@ -0,0 +1,115 @@
+# d1_4_visualization_coffee.py
+# Day 1: 資料科學與數據視覺化 - 咖啡廳營運數據進階分析（學員填空）
+# cd examples/Day1/student && python3 d1_4_visualization_coffee.py
+# 參考答案：../Teacher/d1_4_visualization_coffee.py
+
+import os
+import pandas as pd
+import matplotlib.pyplot as plt
+
+
+def load_csv(filename="coffee_sales_data.csv"):
+    """安全載入 CSV 的輔助函式，用以防範在不同執行路徑下相對路徑讀檔失敗的問題"""
+    # === 【學員填空練習點 1：CSV 檔案安全讀取】 ===
+    # TODO: 1. 使用 os.path.join 將當前腳本所在的目錄與檔案名稱拼接成絕對路徑
+    #       2. 使用 os.path.exists 判斷路徑是否存在，若存在則用 pd.read_csv 讀取並回傳
+    path = os.path.join(os.path.dirname(__file__), filename)
+    if os.path.exists(path):
+        return pd.read_csv(path)
+    else:
+        print("❌ 找不到 CSV 檔案！請確認 coffee_sales_data.csv 與本腳本在同一目錄。")
+        exit(1)
+
+
+df = load_csv()
+
+# 設定中文字型與負號顯示防禦，避免圖表出現亂碼方塊
+plt.rcParams['font.sans-serif'] = [
+    'PingFang TC', 'Heiti TC', 'Arial Unicode MS',
+    'Microsoft JhengHei', 'Noto Sans CJK TC', 'SimHei',
+]
+plt.rcParams['axes.unicode_minus'] = False
+
+# === 【學員填空練習點 2：建立子圖畫布 (Subplots)】 ===
+# TODO: 使用 plt.subplots 建立一個 1 列 2 行 (1x2) 的子圖畫布，尺寸大小為寬 15 英吋、高 6 英吋
+#       並指派給 fig 與 (ax1, ax2)
+fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(15, 6))
+
+
+# ==========================================
+# 1. 子圖 1 (ax1)：營業額與咖啡銷量趨勢對比 (雙 Y 軸)
+# ==========================================
+# === 【學員填空練習點 3：繪製雙 Y 軸折線圖】 ===
+# TODO: 
+# 1. 在 ax1 (左軸) 繪製「營業額」折線圖：X軸為 df['Date'], Y軸為 df['Sales_Amount']
+#    設定參數：點形 marker='o', 顏色 color='#b87e6b' (鐵鏽紅), 圖例 label='營業額 (NTD)', 線寬 linewidth=2
+# 2. 呼叫 ax1.twinx() 建立共用 X 軸的右側 Y 軸物件，指派給 ax1_twin
+# 3. 在 ax1_twin (右軸) 繪製「咖啡銷量」折線圖：X軸為 df['Date'], Y軸為 df['Coffee_Cups_Sold']
+#    設定參數：點形 marker='s' (正方形), 顏色 color='#5f7186' (石板藍), 虛線 linestyle='--', 圖例 label='咖啡銷量 (杯)'
+# 4. 分別設定左軸與右軸的 Y 軸標籤與顏色以示區隔 (提示：使用 set_ylabel 與 tick_params)
+
+# 請在下方完成 ax1 雙 Y 軸折線圖的繪製：
+line1 = ax1.plot(df['Date'], df['Sales_Amount'], marker='o', color="#640f91", label='营业额 (NTD)', linewidth=2 )
+ax1_twin = ax1.twinx()
+line2 = ax1_twin.plot(df['Date'], df['Coffee_Cups_Sold'], marker='s', color= "#19077D", linestyle='--', label='咖啡销量 (杯)', linewidth=2)
+
+ax1.set_ylabel("营业额 (NTD)", color='#640f91', fontsize=12)
+ax1_twin.set_ylabel("咖啡销量 (杯)", color='#19077D', fontsize=12)
+ax1.tick_params(axis='y', labelcolor='#640f91')
+ax1_twin.tick_params(axis='y', labelcolor='#19077D')
+
+# === 【學員填空練習點 4：雙 Y 軸圖例合併】 ===
+# TODO:
+# 1. 由於雙 Y 軸屬於不同的 Axes 物件，若直接呼叫 ax1.legend() 只會顯示左軸的圖例。
+#    請獲取兩軸的 lines 與 labels，合併後統一顯示在 ax1 的左上角 (loc='upper left')
+#    提示：
+#    lines1, labels1 = ax1.get_legend_handles_labels()
+#    lines2, labels2 = ax1_twin.get_legend_handles_labels()
+#    ax1.legend(lines1 + lines2, labels1 + labels2, loc='upper left')
+
+# 請在下方完成圖例合併：
+lines = line1 + line2
+labels = [l.get_label() for l in lines]
+ax1.legend(lines, labels, loc='upper left')
+
+# 設定 ax1 基礎格式
+ax1.set_title("咖啡廳每日營業額與咖啡銷量趨勢", fontsize=14, fontweight='bold')
+ax1.set_xlabel("日期")
+ax1.grid(True, linestyle='--', alpha=0.5)
+ax1.tick_params(axis='x', rotation=45)
+
+
+# ==========================================
+# 2. 子圖 2 (ax2)：促銷活動對客流量的影響 (數據聚合長條圖)
+# ==========================================
+# === 【學員填空練習點 5：數據聚合與長條圖繪製】 ===
+# TODO:
+# 1. 使用 groupby 依據 'Promotion_Active' 欄位分組，並計算 'Customer_Count' 的平均值 (mean)
+#    提示：promo_avg = df.groupby('Promotion_Active')['Customer_Count'].mean()
+# 2. 在 ax2 上繪製長條圖，以 promo_avg.index 作為 X 軸，promo_avg.values 作為 Y 軸
+#    設定參數：顏色 color=['#e6e2d8', '#b87e6b']，外框顏色 edgecolor='#3d3a36'，寬度 width=0.5
+# 3. 在長條圖上方顯示具體的平均數值 (提示：使用 ax2.bar_label 或是跑迴圈加文字)
+
+# 請在下方完成資料聚合與長條圖繪製：
+promo_avg = df.groupby('Promotion_Active')['Customer_Count'].mean()
+bars = ax2.bar(promo_avg.index, promo_avg.values, color=["#15c81e", "#1a54f3"], edgecolor="#bd10cd", width=0.5)
+ax2.bar_label(bars,fmt='%.1f 人', label_type='edge', padding=3, fontweight='bold')
+
+# 設定 ax2 基礎格式
+ax2.set_title("促銷活動對每日平均客流量的影響", fontsize=14, fontweight='bold')
+ax2.set_xlabel("是否舉辦促銷活動")
+ax2.set_ylabel("平均來客數 (人)")
+ax2.grid(True, axis='y', linestyle='--', alpha=0.5)
+
+
+plt.tight_layout()  # 自動調整間距，避免文字重疊
+
+# === 【學員填空練習點 6：圖表儲存與顯示】 ===
+# TODO: 1. 呼叫 savefig 將畫布儲存為檔名 "coffee_sales_analysis.png"，並設定 dpi=300
+#       2. 呼叫 show 顯示圖表
+#       (⚠️ 注意：不要在 print 或 savefig 內使用不支援 cp950 的 emoji)
+
+output_image = "coffee_sales_analysis.png"
+# 請在下方完成圖表存檔與顯示：
+plt.savefig(output_image, dpi=300)
+plt.show()

projects/2026_AI營1/劉宗穎/嘴巴侦测/step3_mouth_ratio.py

@@ -0,0 +1,113 @@
+"""
+Week 11 - Step 2: Mouth Opening Ratio Detection (Template 填空版)
+=============================================================
+本步驟目標：
+  1. 找出特定的臉部特徵點：眼角（33, 263）與嘴唇（13, 14）
+  2. 計算兩眼距離，當作臉寬基準比例尺，防止人靠近或遠離鏡頭造成的誤判 (歸一化)
+  3. 計算嘴巴張開比例：嘴巴距離 / 兩眼距離
+  4. 在 OpenCV 畫面上即時顯示比例與嘴巴開合狀態 (OPEN / CLOSED)
+"""
+import cv2
+import mediapipe as mp
+import math
+import sys
+import os
+
+# 1. 初始化攝影機
+cap = cv2.VideoCapture(0)
+if not cap.isOpened():
+    print("Error: Could not open camera.")
+    sys.exit()
+
+# 2. 載入 MediaPipe 臉部偵測器模型 (Face Landmarker)
+model_path = os.path.join('face_landmarker.task')
+if not os.path.exists(model_path):
+    print(f"Error: Model file not found at {model_path}")
+    sys.exit()
+
+try:
+    from mediapipe.tasks import python
+    from mediapipe.tasks.python import vision
+    base_options = python.BaseOptions(model_asset_path=model_path)
+    options = vision.FaceLandmarkerOptions(base_options=base_options, num_faces=1)
+    detector = vision.FaceLandmarker.create_from_options(options)
+    print("Face Landmarker initialized successfully.")
+except Exception as e:
+    print(f"Error: Failed to initialize detector: {e}")
+    sys.exit()
+
+# 特徵點索引定義
+LEFT_EYE_OUTER = 33
+RIGHT_EYE_OUTER = 263
+LIP_TOP = 13
+LIP_BOTTOM = 14
+
+# 嘴巴張開的比例閾值 (超過代表張開)
+THRESHOLD = 0.25
+
+print("Press 'q' in the window to exit.")
+
+# 3. 主迴圈
+while True:
+    success, img = cap.read()
+    if not success:
+        break
+
+    img = cv2.flip(img, 1)
+    img_h, img_w, _ = img.shape
+
+    # 影像處理
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+    results = detector.detect(mp_image)
+
+    if results.face_landmarks:
+        face_lms = results.face_landmarks[0]
+
+        # 【TODO 1】取得這 4 個特徵點在畫面中的 (x, y) 像素座標，並存在 pts 字典中
+        pts = {}
+        for idx in [LEFT_EYE_OUTER, RIGHT_EYE_OUTER, LIP_TOP, LIP_BOTTOM]:
+            pt = face_lms[idx]
+            pts[idx] = (int(pt.x * img_w), int(pt.y * img_h))
+
+
+        # 【TODO 2】計算兩眼外角距離 (做為基準比例尺)
+        p_lx, p_ly = pts[LEFT_EYE_OUTER]
+        p_rx, p_ry = pts[RIGHT_EYE_OUTER]
+        base_dist = math.hypot(p_rx - p_lx, p_ry - p_ly)
+
+
+        if base_dist > 0:
+            # 【TODO 3】計算嘴巴張開的實際像素距離，並除以眼距算出「嘴巴張開比例 (mouth_ratio)」
+            p_top_x, p_top_y = pts[LIP_TOP]
+            p_bot_x, p_bot_y = pts[LIP_BOTTOM]
+            mouth_dist = math.hypot(p_bot_x - p_top_x, p_bot_y - p_top_y)
+            mouth_ratio = mouth_dist / base_dist
+
+
+            # 【TODO 4】判斷嘴巴狀態是否大於 THRESHOLD，並在畫面上繪製量測輔助線
+            # 提示：用 cv2.line 畫出眼部基準線 (藍色) 和嘴巴測量線 (橘色)
+            is_open = mouth_ratio > THRESHOLD
+            cv2.line(img, (p_lx, p_ly), (p_rx, p_ry), (255, 0, 0), 2)  # 藍色眼線
+            cv2.line(img, (p_top_x, p_top_y), (p_bot_x, p_bot_y), (0, 165, 255), 2)  # 橘色嘴線
+
+
+            # 畫出特徵點小黃圓圈
+            for idx, pos in pts.items():
+                cv2.circle(img, pos, 5, (0, 255, 255), -1)
+
+            # 顯示比例與開合狀態文字
+            cv2.putText(img, f"Mouth Ratio: {mouth_ratio:.2f}", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 255, 255), 2)
+            status_text = "OPEN" if is_open else "CLOSED"
+            status_color = (0, 255, 0) if is_open else (0, 0, 255)
+            cv2.putText(img, f"State: {status_text}", (10, 60), cv2.FONT_HERSHEY_SIMPLEX, 0.7, status_color, 2)
+    else:
+        cv2.putText(img, "No Face Detected", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
+
+    cv2.imshow("Mouth Ratio Detector - Press 'q' to quit", img)
+
+    if cv2.waitKey(1) & 0xFF == ord('q'):
+        break
+
+cap.release()
+cv2.destroyAllWindows()

projects/2026_AI營1/劉宗穎/天气资料图表/d1_3_visualization.py

@@ -0,0 +1,87 @@
+# d1_3_visualization.py
+# Day 1: 資料科學與數據視覺化 - Matplotlib 繪圖基礎（學員填空）
+# cd examples/Day1/student && python3 d1_3_visualization.py
+# 參考答案：../Teacher/d1_3_visualization.py
+
+import os
+import pandas as pd
+import matplotlib.pyplot as plt
+
+def load_csv(filename="taipei_weather_2026.csv"):
+    """安全載入 CSV 的輔助函式，用以防範在不同執行路徑下相對路徑讀檔失敗的問題"""
+    # === 【學員填空練習點 1：CSV 檔案安全讀取】 ===
+    # TODO: 1. 使用 os.path.join 將當前腳本所在的目錄與檔案名稱拼接成絕對路徑
+    #          提示: 當前腳本目錄為 os.path.dirname(__file__)
+    #       2. 使用 os.path.exists 判斷路徑是否存在，若存在則用 pd.read_csv 讀取並回傳
+    path = os.path.join(os.path.dirname(__file__), filename)
+
+    if os.path.exists(path):
+        return pd.read_csv(path)
+    else:
+        print("error: no files found")
+
+df = load_csv()
+
+# 設定中文字型與負號顯示防禦，避免圖表出現亂碼方塊
+plt.rcParams['font.sans-serif'] = [
+    'PingFang TC', 'Heiti TC', 'Arial Unicode MS',
+    'Microsoft JhengHei', 'Noto Sans CJK TC', 'SimHei',
+]
+plt.rcParams['axes.unicode_minus'] = False
+
+# === 【學員填空練習點 2：建立子圖畫布 (Subplots)】 ===
+# TODO: 使用 plt.subplots 建立一個 2 列 1 行 (2x1) 的子圖畫布，尺寸大小為寬 10 英吋、高 8 英吋
+#       並指派給 fig 與 (ax1, ax2)
+fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(10, 8))
+
+# 1. 繪製折線圖 (Line Plot) - 每日溫度
+# === 【學員填空練習點 3：繪製每日溫度折線圖】 ===
+# TODO: 1. 在 ax1 畫布上以 ax1.plot() 繪製折線圖，X軸傳入 df['Date']，Y軸傳入 df['Temperature']
+#          設定參數：帶有圓點 marker='o'，顏色 color='crimson'，圖例 label='溫度 (°C)'，線寬 linewidth=2
+#       2. 使用 ax1.set_title 設定標題 "台北觀測站每日溫度趨勢圖"
+#       3. 設定 ax1.set_xlabel("日期") 與 ax1.set_ylabel("數值")
+#       4. 設定網格 ax1.grid(True, linestyle='--', alpha=0.5)
+#       5. 呼叫 ax1.legend() 顯示圖例
+#       6. 設定 X軸刻度旋轉 ax1.tick_params(axis='x', rotation=45)
+#
+# 請在下方完成 ax1 的溫度折線圖繪製：
+ax1.plot(df['Date'], df['Temperature'], marker='o', color= 'crimson',label='温度 (°C)', linewidth=2)
+ax1.set_title("台北观测站每日温度趋势图", fontsize=14, fontweight='bold')
+ax1.set_xlabel("日期")
+
+ax1.set_ylabel("数值")
+ax1.grid(True,linestyle="--", alpha=0.5)
+ax1.legend()
+ax1.tick_params(axis='x', rotation=45)
+ax1.plot(df['Date'], df['Humidity'], marker='o', color='lime', label='湿度 (%)', linewidth=2)
+ax1.legend()
+
+# 2. 繪製條形圖 (Bar Plot) - 每日降雨量
+# === 【學員填空練習點 4：繪製每日降雨量長條圖】 ===
+# TODO: 1. 在 ax2 畫布上以 ax2.bar() 繪製長條圖，X軸傳入 df['Date']，Y軸傳入 df['Rainfall']
+#          設定參數：顏色 color='skyblue'，外框顏色 edgecolor='royalblue'，圖例 label='降雨量 (mm)'
+#       2. 使用 ax2.set_title 設定標題 "台北觀測站每日降雨量分佈"
+#       3. 設定 ax2.set_xlabel("日期") 與 ax2.set_ylabel("降雨量 (mm)")
+#       4. 設定 Y 軸網格 ax2.grid(True, axis='y', linestyle='--', alpha=0.5)
+#       5. 呼叫 ax2.legend() 顯示圖例
+#       6. 設定 X軸刻度旋轉 ax2.tick_params(axis='x', rotation=45)
+#
+# 請在下方完成 ax2 的降雨量長條圖繪製：
+ax2.bar(df['Date'], df['Rainfall'], color='turquoise',edgecolor = 'turquoise',label='降雨量 (mm)')
+ax2.set_title("台北每天降雨量长条图", fontsize=14, fontweight='bold')
+ax2.set_xlabel("日期")
+ax2.set_ylabel("降雨量")
+ax2.grid(True, axis='y', linestyle='--', alpha=0.5)
+ax2.legend()
+ax2.tick_params(axis='x', rotation=45)
+
+plt.tight_layout()  # 自動調整間距，避免文字重疊
+
+# === 【學員填空練習點 5：圖表自動儲存與顯示】 ===
+# TODO: 1. 呼叫 plt.savefig 將畫布儲存為檔名 "taipei_weather_2026.png"，並設定高畫質 dpi=300
+#          (⚠️ 注意：savefig 必須在 show() 之前呼叫)
+#       2. 呼叫 plt.show() 將圖表視窗顯示在畫面上
+#
+# 請在下方完成圖表存檔與顯示：
+plt.savefig("taipei_weather_2026.png", dpi=300)
+plt.show()

projects/2026_AI營1/劉宗穎/天气资料图表/taipei_weather_2026.csv

@@ -0,0 +1,16 @@
+Date,Temperature,Humidity,Rainfall
+2026-07-01,28.5,82,5.2
+2026-07-02,29.0,80,0.0
+2026-07-03,30.2,75,0.0
+2026-07-04,31.5,70,0.0
+2026-07-05,27.8,90,25.4
+2026-07-06,26.5,95,42.0
+2026-07-07,28.0,85,12.5
+2026-07-08,29.5,78,0.0
+2026-07-09,30.8,72,0.0
+2026-07-10,32.0,68,0.0
+2026-07-11,28.2,88,18.0
+2026-07-12,27.0,92,30.5
+2026-07-13,29.2,80,2.1
+2026-07-14,30.0,76,0.0
+2026-07-15,31.2,71,0.0

projects/2026_AI營1/劉宗穎/影像处理/d3_1_opencv_basic.py

@@ -0,0 +1,70 @@
+# d3_1_opencv_basic.py
+# Day 3: 電腦視覺與影像辨識 - OpenCV 影像處理基礎 (學員練習版)
+# 練習指引：依序找到程式碼中的 TODO 1 到 TODO 5，將其下方的程式碼「解除註釋（刪除行首的 # ）」並執行，觀察影像處理管線中每一步的變化。
+
+import cv2
+import numpy as np
+
+# 1. 影像讀取 (Read)
+img = cv2.imread("onepiece(2).jpeg")
+if img is None:
+    print("[錯誤] 無法讀取 jpg")
+    exit(1)
+
+# 預設灰階影像為原始彩色影像，避免未解開註釋時程式出錯
+gray = img
+
+# === 【重要觀念 TODO 1：灰階轉換 (Grayscale Conversion)】 ===
+# 觀念重點：大多數電腦視覺演算法在處理單通道（黑白）影像時速度更快。
+# 將 BGR 彩色圖像轉換為單通道灰階圖像，可以減少 2/3 的資料量，且能過濾掉無用的顏色雜訊。
+# 請解開下方註釋 (刪除行首的 #) 並執行，觀察變為灰階後的影像。
+gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+
+# 預設模糊影像為灰階影像，避免未解開註釋時程式出錯
+blurred = gray
+
+# === 【重要觀念 TODO 2：高斯模糊平均來平 (Gaussian Blur)】 ===
+# 觀念重點：高斯模糊是透過像素加權滑影像，主要目的是「消除噪點和細微反光」。
+# 這可以避免後續邊緣偵測時把細微的噪輪廓，讓偵測出來的邊緣更加平滑。
+# 請解開下方註釋 (刪除行首的 #) 並執行，觀察影像邊緣的柔焦效果。
+blurred = cv2.GaussianBlur(gray, (5, 5), 0)
+
+# 預設邊緣影像為全黑畫布，避免未解開註釋時點誤判為程式出錯
+edges = np.zeros(img.shape[:2], dtype=np.uint8)
+
+# === 【重要觀念 TODO 3：Canny 邊緣偵測 (Canny Edge Detection)】 ===
+# 觀念重點：Canny 演算法會找出影像中明暗變化劇烈的分界線。它使用雙閥值法：
+# 高於高閥值(150)的 100% 判定為邊緣；介於高低閥值(50~150)之間的若連接強邊緣則保留；其餘捨棄。
+# 請解開下方註釋 (刪除行首的 #) 並執行，觀察提取出的白色線條輪廓。
+edges = cv2.Canny(blurred, 10, 80)
+
+# 預設無偵測到輪廓，且繪製圖像為原始圖像，避免未解開註釋時程式出錯
+contours = []
+contour_img = img.copy()
+
+# === 【重要觀念 TODO 4：尋找與繪製輪廓 (Find & Draw Contours)】 ===
+# 觀念重點：`findContours` 會在二值化(邊緣)影像中尋找相連的邊界，並建立幾何結構。
+# `drawContours` 則會將這些幾何結構繪製到指定的影像上（此處使用綠色框繪製於原始彩色影像）。
+# 請解開下方所有註釋 (刪除行首的 #) 並執行，觀察圖形外框被綠線框起來的結果。
+contours, hierarchy = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+print(f"在影像中偵測到 {len(contours)} 個獨立輪廓特徵。")
+contour_img = img.copy()
+cv2.drawContours(contour_img, contours, -1, (173,216,203), 1)
+alpha=1
+
+# === 【重要觀念 TODO 5：視窗顯示與釋放 (Window Display)】 ===
+# 觀念重點：使用 cv2.imshow 顯示各步驟處理結果，並用 waitKey(0) 讓程式停下等待按鍵，
+# 最後使用 destroyAllWindows 釋放視窗資源。
+# 請解開下方所有註釋 (刪除行首的 #)，以觀看影像處理的成果！
+print("\n正在顯示影像處理結果（請點擊影像視窗並按下任意鍵關閉）...")
+cv2.imshow("Original Shapes", img)
+cv2.imshow("Grayscale", gray)
+cv2.imshow("Canny Edges", edges)
+cv2.imshow("Contours Output", contour_img)
+
+success_gray = cv2.imwrite("output_gray.jpg", gray)
+success_edges = cv2.imwrite("output_edges.jpg", edges)
+success_contour = cv2.imwrite("output_contours.jpg", contour_img)
+
+cv2.waitKey(0)
+cv2.destroyAllWindows()

projects/2026_AI營1/劉宗穎/手指画画游戏/draw_game_student.py

@@ -0,0 +1,167 @@
+import cv2
+import numpy as np
+import mediapipe as mp
+from mediapipe.tasks import python
+from mediapipe.tasks.python import vision
+
+# ============================================
+# Week 9 - 簡單學生版：支援多色切換的 AR 空中寫字板 (純黑畫布版)
+# ============================================
+
+# === 【重要觀念 TODO 1：MediaPipe 手部偵測模型初始化】 ===
+# 觀念重點：
+# 1. BaseOptions 用於設定模型檔案的路徑，這裡使用下載好的 'hand_landmarker.task'。
+# 2. HandLandmarkerOptions 用於設定手部偵測的細項設定，例如 num_hands=1 代表最多偵測一隻手。
+# 3. 最後透過 vision.HandLandmarker.create_from_options 建立手部偵測器物件。
+# 請解開下方三行程式碼的註釋：
+base_options = python.BaseOptions(model_asset_path='hand_landmarker.task')
+options = vision.HandLandmarkerOptions(base_options=base_options, num_hands=1)
+detector = vision.HandLandmarker.create_from_options(options)
+
+cap = cv2.VideoCapture(0)
+canvas = None
+px, py = 0, 0
+
+# 定義幾種高對比的顏色 (格式為 B, G, R)
+COLORS = {     
+    'red': (0, 0, 255),      
+    'orange': (0, 165, 255),
+    'yellow': (0, 255, 255), 
+    'green': (0, 255, 0),
+    'cyan': (255, 255, 0),
+    'dark_blue': (255, 0, 0),
+    'light_purple': (164, 97, 237),
+    'purple': (63, 0, 127),
+    'magenta': (255, 0, 255),
+}
+
+# 預設使用螢光綠
+current_color = COLORS['green']
+
+print("系統啟動中...")
+print("【操作指南】")
+print("☝️  單伸食指：下筆寫字")
+print("✌️  食指+中指：懸停移動 (不寫字)")
+print("⌨️  按下 '1~9'：切換畫筆顏色")
+print("⌨️  按下 'e'：清空畫面")
+print("⌨️  按下 'q'：結束程式")
+
+while True:
+    success, img = cap.read()
+    if not success:
+        break
+
+    # === 【重要觀念 TODO 2：攝影機影像水平翻轉】 ===
+    # 觀念重點：
+    # 1. 攝影機預設取得的影像與我們的真實動作是相反的（非鏡像）。
+    # 2. 在互動式 AR 應用中，為了讓使用者的左手移動時，畫面中的手也往左移動，
+    #    必須使用 cv2.flip(img, 1) 將影像水平翻轉，以提供符合直覺的「鏡像體驗」。
+    # 請解開下方一行程式碼的註釋：
+    img = cv2.flip(img, 1)
+
+    h, w, c = img.shape
+
+    # 初始化畫布
+    if canvas is None:
+        canvas = np.zeros((h, w, c), dtype=np.uint8)
+
+    # 為了維持畫筆提示點（或懸停圓圈）的動態顯示，且不破壞原本的 canvas 紀錄，
+    # 我們在每一訊框都建立一個純黑的暫時畫布來呈現最終畫面。
+    display_img = np.zeros((h, w, c), dtype=np.uint8)
+
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+    results = detector.detect(mp_image)
+
+    if results.hand_landmarks:
+        hand_lms = results.hand_landmarks[0]
+
+        index_tip = hand_lms[8]
+        index_pip = hand_lms[6]
+        middle_tip = hand_lms[12]
+        middle_pip = hand_lms[10]
+
+        cx, cy = int(index_tip.x * w), int(index_tip.y * h)
+
+        index_up = index_tip.y < index_pip.y
+        middle_up = middle_tip.y < middle_pip.y
+
+        if index_up and not middle_up:
+            # 狀態：【繪圖模式】
+            if px == 0 and py == 0:
+                px, py = cx, cy
+
+            # === 【重要觀念 TODO 3：在 OpenCV Canvas 上繪製軌跡】 ===
+            # 觀念重點：
+            # 1. cv2.line 用於在影像上繪製直線。
+            # 2. 參數說明：畫布(canvas)、起點(px, py)、終點(cx, cy)、顏色(current_color)、線寬(8)。
+            # 3. 藉由在每一訊框連續繪製前一個位置到當前位置的線段，即可形成平滑的寫字軌跡。
+            # 請解開下方一行程式碼的註釋：
+            cv2.line(canvas, (px, py), (cx, cy), current_color, 8)
+            
+            px, py = cx, cy
+
+            # 筆頭提示點直接畫在暫時的顯示畫布上，避免在黑畫布上找不到筆尖
+            cv2.circle(display_img, (cx, cy), 10, current_color, cv2.FILLED)
+
+        elif index_up and middle_up:
+            # 狀態：【懸停模式】
+            px, py = 0, 0
+            # 懸停提示圈畫在暫時的顯示畫布上，移動時才不會在黑背景中迷路
+            cv2.circle(display_img, (cx, cy), 12, (255, 255, 255), 2)
+
+        else:
+            px, py = 0, 0
+    else:
+        px, py = 0, 0
+
+    # 為什麼改用 canvas 與 display_img 疊加？
+    # 原本是與相機影像 (img) 疊加，現在改與純黑的 display_img 疊加，
+    # 這樣畫面上就只會留下寫字的軌跡以及當前的筆頭/懸停提示，完全遮蔽人像背景。
+    
+    # === 【重要觀念 TODO 4：AR 畫面疊加與顯示】 ===
+    # 觀念重點：
+    # 1. cv2.add 用於將兩個相同大小與類型的影像矩陣進行相加。
+    # 2. 這裡將寫字痕跡畫布 (canvas) 與當前的筆頭/懸停提示畫布 (display_img) 疊加。
+    # 3. 疊加後使用 cv2.imshow 在視窗中顯示最終成果。
+    # 請解開下方兩行程式碼的註釋：
+    final_output = cv2.add(canvas, display_img)
+    cv2.imshow("Hand Tracking - AR Canvas", final_output)
+
+    # 鍵盤監聽與控制邏輯
+    key = cv2.waitKey(1) & 0xFF
+    if key == ord('q'):
+        break
+    elif key == ord('e'):
+        canvas = np.zeros((h, w, c), dtype=np.uint8)
+
+    elif key == ord('1'):
+        current_color = COLORS['red']
+        print("已切換為：红色")
+    elif key == ord('2'):
+        current_color = COLORS['orange']
+        print("已切換為：橘色")
+    elif key == ord('3'):
+        current_color = COLORS['yellow']
+        print("已切換為：黃色")
+    elif key == ord('4'):
+        current_color = COLORS['green']
+        print("已切換為：綠色")
+    elif key == ord('5'):
+        current_color = COLORS['cyan']
+        print("已切換為：青色")
+    elif key == ord('6'):
+        current_color = COLORS['dark_blue']
+        print("已切換為：深蓝色")
+    elif key == ord('7'):
+        current_color = COLORS['light_purple']
+        print("已切換為：浅紫色")
+    elif key == ord('8'):
+        current_color = COLORS['purple']
+        print("已切換為：紫色")
+    elif key == ord('9'):
+        current_color = COLORS['magenta']
+        print("已切換為：粉色")
+
+cap.release()
+cv2.destroyAllWindows()

projects/2026_AI營1/劉宗穎/爬虫分析/d4_1_ex1_books.py

@@ -0,0 +1,94 @@
+# d4_1_ex1_books.py — Day 4 學員填空
+# cd examples/Day4/student && python d4_1_ex1_books.py
+# 參考答案：../d4_1_ex1_books.py
+
+import requests
+from bs4 import BeautifulSoup
+from urllib.parse import urljoin
+
+# 目標網站：專供爬蟲練習的圖書商城沙盒
+URL = "http://books.toscrape.com/"
+
+# === 【學員填空練習點 1：模擬瀏覽器 Header】 ===
+# TODO: 建立 headers 字典，填入 "User-Agent" 來模擬真實瀏覽器（降低被簡易防爬機制拒絕的機率）
+HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) "
+        "Chrome/120.0.0.0 Safari/537.36"
+    )
+}
+
+print(f"正在連線至圖書沙盒網站: {URL} ...")
+
+try:
+    # 發送 GET 請求取得網頁 HTML
+    response = requests.get(URL, headers=HEADERS, timeout=10)
+
+    # 檢查 HTTP 狀態碼是否為 200 (OK)，非 200 時引發 HTTPError 異常
+    response.raise_for_status()
+
+    # 解決編碼問題：強制將解碼設定為 utf-8 (避免 requests 預設 ISO-8859-1 解碼導致英鎊符號亂碼且在 Windows 印出時崩潰)
+    response.encoding = response.apparent_encoding
+
+    print("連線成功！開始解析書籍資料...\n")
+
+    # 初始化 BeautifulSoup 解析器
+    soup = BeautifulSoup(response.text, "html.parser")
+
+    # === 【學員填空練習點 2：定位 HTML 標籤區塊】 ===
+    # TODO: 找出網頁中所有書籍的容器。每本書都被包在一個 <article> 標籤中，且該標籤的 class 固定為 "product_pod"
+    books = soup.find_all("article", class_="product_pod")
+
+    print(f"成功在本頁找到 {len(books)} 本書籍：")
+    print("=" * 70)
+
+    for idx, book in enumerate(books, 1):
+        # === 【學員填空練習點 3：提取書籍名稱與連結】 ===
+        # TODO:
+        # 1. 尋找 <h3> 底下的 <a> 標籤
+        # 2. 取得 <a> 的 "title" 屬性（因為畫面上長度限制，.get_text() 可能會被 ellipsis 截斷）
+        # 3. 取得 <a> 的 "href" 屬性並使用 urljoin 拼接成完整網址
+        title_link = book.find("h3").find("a")
+        title = title_link.get("title")  # 取得屬性 'title' 的值，避免拿到被截斷的文字
+
+        relative_url = title_link.get("href")
+        detail_url = urljoin(URL, relative_url)
+
+        # === 【學員填空練習點 4：提取書籍價格】 ===
+        # 提示：為了避免 Windows (CP950) 終端機不支援英鎊符號 '£' 而編碼崩潰，請加上 .replace("£", "GBP") 將其替換
+        # 觀察：價格存放在 class 為 "price_color" 的 <p> 標籤中，格式如 "£51.77"
+        # 技巧：將英鎊符號 '£' 替換為 'GBP'，避免在 Windows (CP950) 終端機印出時因不支援該字元而崩潰
+        price_text = book.find("p", class_="price_color").get_text().replace("£", "GBP")
+
+        # === 【學員填空練習點 5：提取評價星等 (Rating)】 ===
+        # TODO:
+        # 1. 尋找 class 為 "star-rating" 的 <p> 標籤
+        # 2. 取得該標籤的 "class" 屬性值（這會是一個串列，如 ['star-rating', 'Three']）
+        # 3. 提取該串列的第二個元素作為星等字串 (若存在)
+
+        rating_tag = book.find("p", class_="star-rating")
+        rating_classes = rating_tag.get("class")
+        rating = "無"
+        if len(rating_classes) > 1:
+            rating = rating_classes[1]  # 提取例如 'Three' (代表三顆星)
+
+        # === 【學員填空練習點 6：提取庫存狀態】 ===
+        # TODO: 尋找 class 為 "instock availability" 的 <p> 標籤，取得文字後去除首尾空白 (.strip())
+        stock_text = book.find("p", class_="instock availability").get_text().strip()
+
+        # 印出解析結果
+        print(f"【編號 {idx}】")
+        print(f"[書名] {title}")
+        print(f"[價格] {price_text}")
+        print(f"[評價] {rating} 星")
+        print(f"[狀態] {stock_text}")
+        print(f"[連結] {detail_url}")
+        print("-" * 70)
+
+except requests.exceptions.HTTPError as http_err:
+    print(f"[錯誤] HTTP 連線異常：{http_err}")
+except requests.exceptions.RequestException as e:
+    print(f"[錯誤] 網路請求失敗：{e}")
+except Exception as ex:
+    print(f"[錯誤] 解析資料時發生非預期錯誤：{ex}")

projects/2026_AI營1/劉宗穎/爬虫分析/d4_1_ex2_ptt.py

@@ -0,0 +1,81 @@
+# d4_1_ex2_ptt.py — Day 4 學員填空
+# cd examples/Day4/student && python d4_1_ex2_ptt.py
+# 參考答案：../d4_1_ex2_ptt.py
+
+import requests
+from bs4 import BeautifulSoup
+from urllib.parse import urljoin
+
+# 目標網站：PTT Python 討論板首頁
+URL = "https://www.ptt.cc/bbs/Python/index.html"
+
+# === 【學員填空練習點 1：模擬瀏覽器 Header】 ===
+# TODO: 填入瀏覽器的 User-Agent，否則 PTT 會拒絕請求並回傳 403 錯誤
+HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) "
+        "Chrome/120.0.0.0 Safari/537.36"
+    )
+}
+
+print(f"正在抓取 PTT Python 板最新文章: {URL} ...")
+
+try:
+    response = requests.get(URL, headers=HEADERS, timeout=10)
+    response.raise_for_status()
+
+    soup = BeautifulSoup(response.text, "html.parser")
+
+    # === 【學員填空練習點 2：定位文章區塊】 ===
+    # TODO: 觀察網頁 HTML 結構，找出所有包裹文章資訊的 <div> 標籤（其 class 名稱為何？）
+    posts = soup.find_all("div", class_="r-ent")
+
+    print(f"\n成功取得 {len(posts)} 筆文章項目（含已刪除與置底公告）：")
+    print("=" * 80)
+
+    for idx, post in enumerate(posts, 1):
+        # === 【學員填空練習點 3：提取文章標題與連結並防錯 (Null Check)】 ===
+        # TODO:
+        # 1. 尋找 class 為 "title" 的 <div> 標籤
+        # 2. 在該 <div> 底下尋找 <a> 標籤 (即文章連結)
+        # 3. ⚠️ 重要思考：若文章已刪除，將不會有 <a> 標籤。請撰寫條件判斷：
+        #    - 若 <a> 標籤存在：提取其文字，並使用 urljoin 拼接 href 屬性得到完整網址。
+        #    - 若 <a> 標籤不存在：直接將 title_div 的文字作為標題，並將網址設為 "無 (文章已刪除)"。
+        title_div = post.find("div", class_="title")
+        title_link = title_div.find("a") if title_div else None
+
+        if title_link:
+            title = title_link.get("title")
+            relative_url = title_link.get("href")
+            post_url = urljoin(URL, relative_url)
+        else:
+            # 文章已刪除
+            title_div = post.find("div", class_="title")
+            title = title_div.get_text().strip() if title_div else "（已刪除）"
+            post_url = "無 (文章已刪除)"
+
+
+        # === 【學員填空練習點 4：提取推文數】 ===
+        # TODO: 尋找 class 為 "nrec" 的 <div> 標籤。
+        # 技巧：使用 .get_text().strip() 取得文字。若無文字（0 推），請將變數預設為 "0"。
+        nrec_div = post.find("div", class_="nrec")
+        nrec = nrec_div.get_text().strip() if nrec_div else "0"
+
+        # === 【學員填空練習點 5：提取作者與發文日期】 ===
+        # TODO: 分別尋找 class 為 "author" 與 "date" 的 <div> 標籤，並提取純文字（記得使用 .strip() 去除空白）
+        author = post.find("div", class_="author").get_text().strip()
+        date = post.find("div", class_="date").get_text().strip()
+
+        # 輸出單篇文章資訊
+        print(f"[{idx}] 推文數: {nrec:3s} | 發文日期: {date} | 作者: {author}")
+        print(f"    標題: {title}")
+        print(f"    網址: {post_url}")
+        print("-" * 80)
+
+except requests.exceptions.HTTPError as http_err:
+    print(f"[錯誤] PTT 伺服器拒絕或找不到網頁：{http_err}")
+except requests.exceptions.RequestException as e:
+    print(f"[錯誤] 網路請求失敗：{e}")
+except Exception as ex:
+    print(f"[錯誤] 解析資料時發生非預期錯誤：{ex}")

projects/2026_AI營1/劉宗穎/爬虫分析/d4_1_ex3_ithelp.py

@@ -0,0 +1,98 @@
+# d4_1_ex3_ithelp.py — Day 4 學員填空
+# cd examples/Day4/student && python d4_1_ex3_ithelp.py
+# 參考答案：../d4_1_ex3_ithelp.py
+
+import requests
+from bs4 import BeautifulSoup
+
+# 目標網站：iT 邦幫忙技術文章首頁
+URL = "https://ithelp.ithome.com.tw/articles?tab=tech"
+
+# === 【學員填空練習點 1：模擬瀏覽器 Header】 ===
+# TODO: 填入瀏覽器的 User-Agent，否則可能被伺服器拒絕連線
+HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) "
+        "Chrome/120.0.0.0 Safari/537.36"
+    )
+}
+
+print(f"正在連線至 iT 邦幫忙並抓取技術文章: {URL} ...")
+
+try:
+    response = requests.get(URL, headers=HEADERS, timeout=10)
+    response.raise_for_status()
+
+    print("連線成功！解析文章列表中...\n")
+    soup = BeautifulSoup(response.text, "html.parser")
+
+    # === 【學員填空練習點 2：定位文章卡片區塊】 ===
+    # TODO: 尋找所有包裹文章資訊的 <div> 標籤（其 class 為 "qa-list"）
+    articles = soup.find_all("div", class_="qa-list")
+
+    print(f"成功取得前 {len(articles)} 篇熱門技術文章：")
+    print("=" * 80)
+
+    for idx, art in enumerate(articles, 1):
+        # === 【學員填空練習點 3：提取標題與超連結】 ===
+        # TODO:
+        # 1. 在 art 中尋找 class 為 "qa-list__title-link" 的 <a> 標籤
+        # 2. 提取其純文字內容並去除首尾空白 (.strip())，作為標題
+        # 3. 提取其 "href" 屬性，作為文章連結
+        title_link = art.find("a", class_="qa-list__title-link")
+        if not title_link:
+            continue
+        title = title_link.get_text().strip()
+        post_url = title_link.get("href").strip()
+
+        # === 【學員填空練習點 4：提取作者名稱】 ===
+        # TODO:
+        # 1. 尋找 class 為 "qa-list__info" 的 <div> 標籤
+        # 2. 在該 <div> 下尋找 class 為 "qa-list__info-link" 的 <a> 標籤
+        # 3. 提取其文字內容作為作者名稱
+        info_div = art.find("div", class_="qa-list__info")
+        author = "未知作者"
+        if info_div:
+            author_link = info_div.find("a", class_="qa-list__info-link")
+            if author_link:
+                author = author_link.get_text().strip()
+
+        # === 【學員填空練習點 5：提取發文時間】 ===
+        # TODO: 尋找 class 為 "qa-list__info-time" 的 <a> 標籤，並提取其 "title" 屬性值（因 title 屬性中包含完整時間秒數）
+        time_link = art.find("div", class_="qa-list__info-time")
+        post_time = time_link.get("title") if time_link else "未知時間"
+
+        # === 【學員填空練習點 6：利用 List 索引提取多項數據】 ===
+        # TODO:
+        # 1. 使用 find_all 找出卡片中所有 class 為 "qa-condition" 的 <a> 標籤
+        # 2. 當抓取到的數量大於等於 3 時，利用 List 索引取得對應的資訊：
+        #    - 索引 0 的項目底下的 "qa-condition__count" <span>：Like 數
+        #    - 索引 1 的項目底下的 "qa-condition__count" <span>：回應數
+        #    - 索引 2 的項目底下的 "qa-condition__count" <span>：瀏覽數
+        conditions = art.find_all("a", class_="qa-condition")
+
+        likes = "0"
+        comments = "0"
+        views = "0"
+
+        if len(conditions) >= 3:
+            # 請在此處填入利用 conditions[0]、conditions[1]、conditions[2] 尋找 span 的代碼
+            likes = conditions[0].find("span", class_="qa-condition__count").get_text().strip()
+            comments = conditions[1].find("span", class_="qa-condition__count").get_text().strip()
+            views = conditions[2].find("span", class_="qa-condition__count").get_text().strip()
+
+        # 格式化輸出
+        print(f"【排名 {idx}】")
+        print(f"[標題] {title}")
+        print(f"[作者] {author} | [時間] {post_time}")
+        print(f"[數據] Like: {likes} | 回應: {comments} | 瀏覽/訂閱: {views}")
+        print(f"[連結] {post_url}")
+        print("-" * 80)
+
+except requests.exceptions.HTTPError as http_err:
+    print(f"[錯誤] 伺服器回應異常：{http_err}")
+except requests.exceptions.RequestException as e:
+    print(f"[錯誤] 網路請求失敗：{e}")
+except Exception as ex:
+    print(f"[錯誤] 解析資料時發生非預期錯誤：{ex}")

projects/2026_AI營1/劉宗穎/爬虫分析与AI报告/d4_1_web_scraping.py

@@ -0,0 +1,68 @@
+# d4_1_web_scraping.py — Day4 學員填空
+# cd examples/Day4/student && python3 d4_1_web_scraping.py
+# 參考答案：../d4_1_web_scraping.py
+
+import requests
+from bs4 import BeautifulSoup
+
+# 本範例抓取專供網頁爬蟲教學使用的模擬沙盒網站：Quotes to Scrape
+URL = "http://quotes.toscrape.com/"
+
+# === 【學員填空練習點 1：模擬瀏覽器 Header】 ===
+# TODO: 建立 headers 字典，填入 "User-Agent" 來模擬真實瀏覽器（降低被簡易防爬機制拒絕的機率）
+HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) "
+        "Chrome/120.0.0.0 Safari/537.36"
+    )
+}
+
+print(f"正在連線並抓取網站內容: {URL} ...")
+
+try:
+    # === 【學員填空練習點 2：連線請求與下載網頁】 ===
+    # TODO: 1. 使用 requests.get 發送 GET 請求，帶入 headers=HEADERS 與設定 timeout=5
+    # TODO: 2. 判斷 response.status_code 是否為 200 (下載成功)
+    response = requests.get(URL, headers=HEADERS, timeout=5)
+
+    if response and response.status_code == 200:
+        print("網頁下載成功！開始解析 HTML 結構...")
+
+        # === 【學員填空練習點 3：初始化 BeautifulSoup 解析器】 ===
+        # TODO: 使用 BeautifulSoup 解析 response.text，初始化 html.parser
+        soup = BeautifulSoup(response.text, "html.parser")
+
+        # === 【學員填空練習點 4：定位 HTML 標籤區塊】 ===
+        # TODO: 找出網頁中所有名言與作者的 div 標籤（其 class 為 'quote'）
+        quote_divs = soup.find_all("div", class_="quote")
+
+        print(f"\n成功解析！本日精選名言如下（共 {len(quote_divs)} 筆）：")
+        print("=" * 60)
+
+        # === 【學員填空練習點 5：迴圈遍歷提取內容與防錯】 ===
+        # TODO:
+        # 1. 遍歷 quote_divs，並使用 enumerate 取得索引 idx（從 1 開始）
+        # 2. 用 .find() 尋找 span 且 class 為 'text' 的名言元素
+        # 3. 用 .find() 尋找 small 且 class 為 'author' 的作者元素
+        # 4. 進行 None 防錯處理，確保兩者皆非空值後，使用 .get_text() 提取純文字並印出
+        for idx, quote_div in enumerate(quote_divs, 1):
+            text_el = quote_div.find("span", class_="text")
+            author_el = quote_div.find("small", class_="author")
+            if text_el is None or author_el is None:
+                continue
+            text = text_el.get_text()
+            author = author_el.get_text()
+
+            print(f"[{idx}] \"{text}\"")
+            print(f"    -- by {author}\n")
+
+        print("=" * 60)
+        print("提醒：請尊重 robots.txt 與網站使用條款，勿對目標網站發送高頻請求。")
+    else:
+        status = response.status_code if response else "未知"
+        print(f"[錯誤] 無法抓取網頁，伺服器回應狀態碼：{status}")
+
+except requests.exceptions.RequestException as e:
+    print(f"[錯誤] 連線失敗：{e}")
+    print("請確認電腦的網路連線是否通暢。")

projects/2026_AI營1/劉宗穎/爬虫分析与AI报告/d4_2_agent_task_chain.py

@@ -0,0 +1,141 @@
+# d4_2_agent_task_chain.py
+# Day 4: 自動化任務鏈 (Task Chain / 簡化版 Agent 工作流)
+# 說明：本範例是「固定管線」——步驟由我們寫死，不是模型自主 function calling。
+#       目的是讓學生先理解：工具 + LLM + 記憶狀態 如何串成可交付的自動化流程。
+
+import os
+import requests
+from bs4 import BeautifulSoup
+
+GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-3.1-flash-lite")
+api_key = os.environ.get("GEMINI_API_KEY", "AQ.Ab8RN6JvbUqLW-wcddeMducmta0aGE2QtcrmF4Kfd8nnpZ9mzg")
+
+HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) "
+        "Chrome/120.0.0.0 Safari/537.36"
+    )
+}
+
+
+def plan_steps(topic):
+    """規劃階段：對學生展示任務拆解（固定計畫，非 LLM 自主規劃）。"""
+    steps = [
+        f"1. 使用爬蟲工具擷取主題「{topic}」相關名言",
+        "2. 將原始文字作為 context 送入 LLM 摘要與翻譯",
+        "3. 將 Markdown 報告寫入本機檔案",
+    ]
+    print("[Planning] 任務計畫：")
+    for s in steps:
+        print("  ", s)
+    return steps
+
+
+def tool_scrape_topic_quotes(topic):
+    """
+    工具 A：抓取 Quotes to Scrape 的標籤頁面內容。
+    例如 topic = 'love', 'life', 'inspirational' 等。
+    """
+    url = f"http://quotes.toscrape.com/tag/{topic}/"
+    print(f"[Tool] 執行網頁資料檢索。目標網址: {url}")
+    try:
+        response = requests.get(url, headers=HEADERS, timeout=5)
+        if response.status_code == 200:
+            soup = BeautifulSoup(response.text, "html.parser")
+            quote_divs = soup.find_all("div", class_="quote")
+            results = []
+            for q in quote_divs:
+                text_el = q.find("span", class_="text")
+                author_el = q.find("small", class_="author")
+                if text_el is None or author_el is None:
+                    continue
+                text = text_el.get_text()
+                author = author_el.get_text()
+                results.append(f"名言: {text} | 作者: {author}")
+            return "\n".join(results) if results else "找不到與此主題相關的內容。"
+        return f"網頁讀取錯誤，代碼: {response.status_code}"
+    except Exception as e:
+        return f"工具執行異常: {e}"
+
+
+def tool_save_report(filename, content):
+    """工具 B：將報告寫入本機檔案。"""
+    with open(filename, "w", encoding="utf-8") as f:
+        f.write(content)
+    return filename
+
+
+def run_task_chain(topic):
+    print("=" * 60)
+    print(f"啟動自動化任務鏈。目標：整理與分析關於『{topic}』的文獻報告")
+    print("（模式：固定 Pipeline，非自主 Tool-Calling Agent）")
+    print("=" * 60)
+
+    # 步驟 0: 規劃（教學展示）
+    plan_steps(topic)
+
+    # 步驟 1: 呼叫檢索工具
+    raw_data = tool_scrape_topic_quotes(topic)
+    print(f"\n[步驟 1] 檢索完成！取得原始數據長度: {len(raw_data)} 字元。")
+    print("--- 原始數據預覽 ---")
+    preview = raw_data[:200] + ("..." if len(raw_data) > 200 else "")
+    print(preview)
+    print("--------------------")
+
+    # 工作記憶：保存中間狀態
+    memory = {"topic": topic, "raw_data": raw_data, "report": ""}
+
+    # 步驟 2: LLM 分析（或 Mock 降級）
+    if not api_key:
+        print("\n[步驟 2] 未偵測到 API Key。啟動模擬生成中...")
+        report_content = (
+            f"# 模擬關於《{topic}》的名人名言分析報告\n\n"
+            f"## 一、 數據檢索摘要\n"
+            f"本次任務共檢索到數筆關於 {topic} 的語錄。以下為代表性文獻：\n"
+            f"- 模擬數據：\"Love is all, love is new.\" -- John Lennon\n\n"
+            f"## 二、 核心思想分析\n"
+            f"本機模擬分析指出，此主題的語錄主要著眼於人與人之間的連結與心靈啟發。\n"
+            f"*(注意：此為 Mock 離線生成報告，設定 GEMINI_API_KEY 以啟用 Gemini 真實生成。)*\n"
+        )
+    else:
+        print(f"\n[步驟 2] 正在調用 Gemini ({GEMINI_MODEL}) 進行深度分析與排版...")
+        try:
+            import google.generativeai as genai  # 僅雲端模式需要
+
+            genai.configure(api_key=api_key)
+            model = genai.GenerativeModel(GEMINI_MODEL)
+
+            prompt = (
+                f"請幫我分析以下關於『{topic}』的原始英文名言數據，並用繁體中文撰寫一份 Markdown 報告。\n"
+                f"報告必須包含三個部分：\n"
+                f"1. 導言：說明這個主題的核心意義。\n"
+                f"2. 名言分析：挑選原始資料中 2-3 句具代表性的句子進行翻譯與深度賞析。\n"
+                f"3. 結論：對學生的啟發與建議。\n\n"
+                f"=== 原始數據 ===\n"
+                f"{raw_data}"
+            )
+
+            response = model.generate_content(prompt)
+            report_content = response.text
+            print("[步驟 2] 雲端分析完成！")
+        except Exception as e:
+            print(f"[錯誤] AI 分析失敗: {e}。轉為模擬輸出。")
+            report_content = f"# 分析失敗\n\n錯誤詳情：{e}\n"
+
+    memory["report"] = report_content
+
+    # 步驟 3: 自動存檔
+    output_filename = f"agent_report_{topic}.md"
+    try:
+        saved = tool_save_report(output_filename, report_content)
+        print(f"\n[步驟 3] 任務報告生成成功！已儲存為檔案: {saved}")
+        print("=" * 60)
+    except Exception as e:
+        print(f"[錯誤] 儲存報告失敗: {e}")
+
+
+if __name__ == "__main__":
+    # 可更換為: 'love', 'life', 'inspirational', 'humor' 等
+    test_topic = "s"
+    run_task_chain(test_topic)

projects/2026_AI營1/劉宗穎/爬虫分析与AI报告/d4_2_ex2_crawler_ai.py

@@ -0,0 +1,231 @@
+# d4_2_ex2_crawler_ai.py
+# Day 4: 爬蟲與 AI 分析任務鏈 (學生填空版)
+# 執行方式：cd examples/Day4/student && python d4_2_ex2_crawler_ai.py
+
+import os
+import requests
+from bs4 import BeautifulSoup
+from urllib.parse import urljoin
+from google import genai
+
+# 設定預設使用的 Gemini 模型
+GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-3.1-flash-lite")
+api_key = os.environ.get("GEMINI_API_KEY", "AQ.Ab8RN6JvbUqLW-wcddeMducmta0aGE2QtcrmF4Kfd8nnpZ9mzg")
+
+HEADERS = {
+    "User-Agent": (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+        "AppleWebKit/537.36 (KHTML, like Gecko) "
+        "Chrome/120.0.0.0 Safari/537.36"
+    )
+}
+
+def plan_steps(source_type):
+    """規劃階段：展示任務拆解流程。"""
+    steps = [
+        f"步驟 1: 啟動爬蟲工具擷取『{source_type.upper()}』熱門文章與輿情數據",
+        "步驟 2: 將爬取資料作為 Context，呼叫 Gemini 進行輿情趨勢與熱門話題分析",
+        "步驟 3: 自動將 AI 生成的繁體中文分析報告以 UTF-8 編碼存為 Markdown 檔案"
+    ]
+    print("\n[Planning] 任務計畫已啟動：")
+    for s in steps:
+        print(f"  -> {s}")
+    return steps
+
+def tool_scrape_ptt():
+    """工具 A：抓取 PTT Python 板最新文章資訊。"""
+    url = "https://www.ptt.cc/bbs/Tennis/index.html"
+    print(f"\n[Tool] 正在爬取 PTT Python 板文章: {url} ...")
+    try:
+        response = requests.get(url, headers=HEADERS, timeout=10)
+        response.raise_for_status()
+
+        soup = BeautifulSoup(response.text, "html.parser")
+        posts = soup.find_all("div", class_="r-ent")
+
+        results = []
+        for post in posts:
+
+            # === 【學員填空練習點 1：PTT 文章標題與連結提取及防錯 (Null Check)】 ===
+            # TODO 1: 若文章已被作者或板主刪除，title_div 底下將會「沒有 <a> 標籤」
+            # 1. 請在 title_div 底下尋找 <a> 標籤 (指派給 title_link 變數)
+            # 2. 撰寫條件判斷：若 title_link 存在，提取其純文字並用 urljoin 拼接成完整網址。
+            #    若不存在，直接將 title_div 的純文字作為 title，且將連結 post_url 設為 "無 (文章已刪除)"。
+
+                # 請在此處實作 TODO 1 的邏輯...
+            title_div = post.find("div", class_="title")
+            title_link = title_div.find("a") if title_div else None
+
+            if title_link:
+                title = title_link.get_text().strip()
+                relative_url = title_link.get("href")
+                post_url = urljoin(url, relative_url)
+            else:
+                # 文章已刪除
+                title_div = post.find("div", class_="title")
+                title = title_div.get_text().strip() if title_div else "（已刪除）"
+                post_url = "無 (文章已刪除)"
+
+            nrec_div = post.find("div", class_="nrec")
+            nrec = nrec_div.get_text().strip() if nrec_div else "0"
+            if not nrec:
+                nrec = "0"
+
+            author = post.find("div", class_="author").get_text().strip() if post.find("div", class_="author") else "未知作者"
+            date = post.find("div", class_="date").get_text().strip() if post.find("div", class_="date") else "未知日期"
+
+            results.append(f"標題: {title} | 作者: {author} | 日期: {date} | 推文數: {nrec} | 網址: {post_url}")
+
+        return "\n".join(results) if results else "找不到任何文章項目。"
+    except Exception as e:
+        return f"[PTT 爬蟲錯誤] 執行異常: {e}"
+
+def tool_scrape_ithelp():
+    """工具 B：抓取 iT 邦幫忙熱門技術文章。"""
+    url = "https://ithelp.ithome.com.tw/articles?tab=tech"
+    print(f"\n[Tool] 正在爬取 iT 邦幫忙技術文章: {url} ...")
+    try:
+        response = requests.get(url, headers=HEADERS, timeout=10)
+        response.raise_for_status()
+
+        soup = BeautifulSoup(response.text, "html.parser")
+        articles = soup.find_all("div", class_="qa-list")
+
+        results = []
+        for art in articles:
+            title_link = art.find("a", class_="qa-list__title-link")
+            if not title_link:
+                continue
+            title = title_link.get_text().strip()
+            post_url = title_link.get("href").strip()
+
+            info_div = art.find("div", class_="qa-list__info")
+            author = "未知作者"
+            if info_div:
+                author_link = info_div.find("a", class_="qa-list__info-link")
+                if author_link:
+                    author = author_link.get_text().strip()
+
+            time_link = art.find("a", class_="qa-list__info-time")
+            post_time = time_link.get("title") if time_link else "未知時間"
+
+            conditions = art.find_all("a", class_="qa-condition")
+            likes, comments, views = "0", "0", "0"
+
+            # === 【學員填空練習點 2：利用 List 索引提取多項數據】 ===
+            # TODO 2: 當抓取到的 conditions 數量大於等於 3 時，利用 List 索引取得對應的資訊：
+            # - 索引 0 的項目底下的 "qa-condition__count" <span>：Like 數
+            # - 索引 1 的項目底下的 "qa-condition__count" <span>：回應數
+            # - 索引 2 的項目底下的 "qa-condition__count" <span>：瀏覽數
+            if len(conditions) >= 3:
+                # 請在下方填入提取 Like、回應與瀏覽數的代碼：
+                likes = conditions[0].find("span", class_="qa-condition__count").get_text().strip()
+            comments = conditions[1].find("span", class_="qa-condition__count").get_text().strip()
+            views = conditions[2].find("span", class_="qa-condition__count").get_text().strip()
+
+            results.append(
+                f"標題: {title} | 作者: {author} | 時間: {post_time} | "
+                f"Like: {likes} | 回應: {comments} | 瀏覽: {views} | 網址: {post_url}"
+            )
+
+        return "\n".join(results) if results else "找不到任何技術文章。"
+    except Exception as e:
+        return f"[iT邦幫忙爬蟲錯誤] 執行異常: {e}"
+
+def tool_save_report(filename, content):
+    """工具 C：將報告以 UTF-8 編碼寫入本機檔案。"""
+    print(f"\n[Tool] 正在將分析報告儲存至檔案: {filename} ...")
+    # === 【學員填空練習點 6：寫入檔案編碼處理】 ===
+    # TODO 6: 請使用 with open 開啟檔案，以 "utf-8" 編碼寫入 content，並回傳 filename。
+    # 提示：必須傳入 encoding="utf-8" 以避免 Windows 系統的中文亂碼
+    with open(filename, "w", encoding="utf-8") as f:
+        f.write(content)
+    # 請在此處實作 TODO 6 ...
+    return filename
+
+def run_analysis_chain(source_type):
+    print("=" * 70)
+    print(f"啟動自動化輿情任務鏈。目標來源：『{source_type.upper()}』數據分析")
+    print("=" * 70)
+
+    # 步驟 0: 展示規劃
+    plan_steps(source_type)
+
+    # 步驟 1: 執行對應的爬蟲工具
+    if source_type.lower() == "ptt":
+        raw_data = tool_scrape_ptt()
+    elif source_type.lower() == "ithelp":
+        raw_data = tool_scrape_ithelp()
+    else:
+        print("[錯誤] 未支援的數據來源。僅支援 'ptt' 或 'ithelp'")
+        return
+
+    print(f"\n[步驟 1] 資料檢索完成！取得原始數據長度: {len(raw_data)} 字元。")
+    print("--- 數據預覽 (前 200 字) ---")
+    print(raw_data[:200] + ("..." if len(raw_data) > 200 else ""))
+    print("----------------------------")
+
+    # === 【學員填空練習點 3：任務鏈工作記憶與態狀維護】 ===
+    # TODO 3: 請建立名為 memory 的字典 (dict)，包含三個鍵值:
+    # "source" 指向來源 source_type、"raw_data" 指向原始爬取資料 raw_data、"report" 設為空字串 "" 用以儲存後續 AI 報告。
+    memory = {
+        "source": source_type,
+        "raw_data": raw_data,
+        "report": ""
+    }
+
+    # 步驟 2: AI 分析 (或 Mock 降級防禦)
+    if not api_key:
+        print("\n[步驟 2] 未偵測到環境變數 GEMINI_API_KEY，自動啟動 Mock 離線降級機制...")
+        # === 【學員填空練習點 5：Mock 降級防衛機制】 ===
+        # TODO 5: 請設計一段模擬的 Markdown 格式分析報告字串，指派給 report_content。
+        # 報告內容至少需要包含標題與數據摘要，避免程式在無 Key 時報錯中斷。
+        report_content = (
+            f"#模拟与情分析报告 - 数据来源:{source_type.upper()}\n\n"
+        )
+        print("[步驟 2] Mock 報告模擬生成完畢。")
+    else:
+        print(f"\n[步驟 2] 正在調用 Gemini ({GEMINI_MODEL}) 進行數據分析與排版...")
+        try:
+            # === 【學員填空練習點 4：新版 google-genai SDK 初始化與調用】 ===
+            # TODO 4:
+            # 1. 呼叫 genai.Client 並傳入 api_key 來初始化 client 物件。
+            # 2. 使用 client.models.generate_content 發送生成請求。
+            #    傳入參數 model=GEMINI_MODEL，contents 則為準備好的 Prompt。
+            client = genai.Client(api_key=api_key)
+            prompt = (
+                f"請幫我分析以下來自 {source_type.upper()} 的最新網頁抓取文章列表，並用繁體中文撰寫一份 Markdown 网球分析報告。\n"
+                f"報告必須包含以下部分：\n"
+                f"1. 说网球界的名人。\n"
+                f"2. 说网球在哪时候被发明的。\n"
+                f"3. 说网球对身体有什么帮助。\n"
+                f"4. 说一些最精彩的比赛。\n\n"
+                f"=== 原始爬蟲數據 ===\n"
+                f"{raw_data}"
+            )
+
+            # 請在此處實作 TODO 4 ...
+            resopnse = client.models.generate_content(
+                model=GEMINI_MODEL,
+                contents=prompt
+            )
+            report_content = ""
+            print("[步驟 2] 雲端 Gemini 分析完成！")
+        except Exception as e:
+            print(f"[錯誤] AI 分析失敗: {e}，自動轉為錯誤備份輸出。")
+            report_content = f"# 輿情分析失敗\n\n錯誤詳情：{e}\n\n=== 原始數據 ===\n{raw_data}"
+
+    memory["report"] = report_content
+
+    # 步驟 3: 自動存檔
+    output_filename = f"ai_analysis_report_{source_type}.md"
+    try:
+        saved_file = tool_save_report(output_filename, memory["report"])
+        print(f"\n[步驟 3] 任務報告生成成功！已儲存至: {saved_file}")
+        print("=" * 70)
+    except Exception as e:
+        print(f"[錯誤] 儲存報告失敗: {e}")
+
+if __name__ == "__main__":
+    # 可以切換為 "ptt" 或 "ithelp" 進行測試
+    run_analysis_chain("ptt")

projects/2026_AI營1/劉宗穎/眼睛距离侦测/step0_eyeguard_template.py

@@ -0,0 +1,156 @@
+"""
+Week 12 - AI 生活工具箱：護眼大師 (Template 填空版)
+=====================================================
+本週目標：
+  1. 將 AI 視覺技術整合進一個「解決生活痛點」的完整專案
+  2. 使用 MediaPipe 臉部偵測兩眼距離，判斷是否離螢幕太近
+  3. 在 Pygame 畫面上顯示即時狀態與警報
+  4. 加入防呆機制（try-except、None 檢查）
+
+本範例：護眼大師 EyeGuard
+  - 偵測臉部兩眼之間的距離（兩眼距離越大 = 離鏡頭越近）
+  - 超過安全閾值時發出警報文字 + 閃爍效果
+  - 適用場景：打電動、寫作業時防止坐姿太差靠太近
+
+注意：
+  - 確保 face_landmarker.task 檔案和這個腳本放在同一個資料夾。
+"""
+import pygame
+import cv2
+import mediapipe as mp
+from mediapipe.tasks import python
+from mediapipe.tasks.python import vision
+import math
+import sys
+import os
+
+# ============================================
+# 1. 初始化
+# ============================================
+pygame.init()
+pygame.mixer.init()
+WIDTH, HEIGHT = 800, 600
+screen = pygame.display.set_mode((WIDTH, HEIGHT))
+pygame.display.set_caption("Week 12 - 護眼大師 EyeGuard 👁️🛡️")
+clock = pygame.time.Clock()
+
+font = pygame.font.SysFont(None, 48)
+small_font = pygame.font.SysFont(None, 28)
+big_font = pygame.font.SysFont(None, 72)
+
+# ============================================
+# 2. 攝影機與 AI 模組
+# ============================================
+cap = cv2.VideoCapture(0)
+# 沒有攝影機就改成影片：cap = cv2.VideoCapture("backup_face_video.mp4")
+
+# 【TODO 1】初始化 MediaPipe 臉部偵測 (Face Landmarker)
+base_options = python.BaseOptions(model_asset_path=os.path.join('face_landmarker.task'))
+options = vision.FaceLandmarkerOptions(base_options=base_options, num_faces=1)
+detector = vision.FaceLandmarker.create_from_options(options)
+
+# 提示：左眼的 Landmark 索引為 33，右眼為 263
+
+# ============================================
+# 3. 遊戲變數
+# ============================================
+eye_distance = 0
+DANGER_THRESHOLD = 120  # 兩眼像素距離超過此值 = 太近了！
+
+# 【TODO 2】警報狀態變數
+is_too_close = False
+was_too_close = False    # 邊緣觸發
+warning_timer = 0        # 警告閃爍計時器
+
+# 統計數據
+total_warnings = 0
+safe_seconds = 0
+frame_counter = 0
+
+# ============================================
+# 4. 主迴圈
+# ============================================
+running = True
+while running:
+    for event in pygame.event.get():
+        if event.type == pygame.QUIT:
+            running = False
+
+    # --- 讀取攝影機 ---
+    success, img = cap.read()
+    if not success:
+        cap.set(cv2.CAP_PROP_POS_FRAMES, 0)
+        continue
+
+    img = cv2.flip(img, 1)
+    frame_counter += 1
+
+    # 【TODO 3】臉部偵測與兩眼距離計算
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+    results = detector.detect(mp_image)
+    
+    if results.face_landmarks:
+        face_lms = results.face_landmarks[0]
+    #
+    #     # 取得左眼 (33) 和右眼 (263) 座標
+        left_eye = face_lms[33]
+        right_eye = face_lms[263]
+    #
+        lx, ly = int(left_eye.x * WIDTH), int(left_eye.y * HEIGHT)
+        rx, ry = int(right_eye.x * WIDTH), int(right_eye.y * HEIGHT)
+    
+        # 計算兩眼之間的像素距離
+        eye_distance = math.hypot(rx - lx, ry - ly)
+    
+        # 【TODO 4】判斷是否太近
+        is_too_close = eye_distance > DANGER_THRESHOLD
+    
+        # 【TODO 5】邊緣觸發：剛進入危險區才播放警報
+        if is_too_close and not was_too_close:
+            total_warnings += 1
+            warning_timer = 60  # 閃爍 60 幀
+            print(f"⚠️ 第 {total_warnings} 次警告：你離螢幕太近了！")
+        
+        was_too_close = is_too_close
+
+    # --- 繪製畫面 ---
+    # 【TODO 6】根據狀態切換背景顏色
+    if is_too_close:
+        if warning_timer > 0 and warning_timer % 4 < 2:
+            screen.fill((80, 20, 20))  # 閃爍紅色
+        else:
+            screen.fill((50, 15, 15))  # 暗紅色
+        warning_timer = max(0, warning_timer - 1)
+    else:
+        screen.fill((15, 30, 20))  # 安全綠色
+    screen.fill((20, 20, 35))  # 預設背景
+
+    # 【TODO 7】繪製距離量表
+    # 提示：用 pygame.draw.rect 畫一個長條，長度 = eye_distance 的比例
+    # 安全時綠色，危險時紅色
+    if is_too_close:
+        pygame.draw.rect(screen, (255, 80, 80), (50, HEIGHT - 50, min(eye_distance * 2, WIDTH - 100), 20))
+    else:
+        pygame.draw.rect(screen, (80, 200, 100), (50, HEIGHT - 50, min(eye_distance * 2, WIDTH - 100), 20))
+    # 【TODO 8】如果太近，顯示大字警告
+    if is_too_close:
+        warn_text = big_font.render("⚠️ TOO CLOSE!", True, (255, 80, 80))
+        screen.blit(warn_text, (WIDTH//2 - warn_text.get_width()//2, HEIGHT//2 - 40))
+    else:
+        safe_text = font.render("✅ Safe Distance", True, (80, 200, 100))
+        screen.blit(safe_text, (WIDTH//2 - safe_text.get_width()//2, HEIGHT//2 - 20))
+
+    # 顯示數值
+    info = small_font.render(f"Eye Distance: {eye_distance:.0f} px  |  Threshold: {DANGER_THRESHOLD}", True, (150, 150, 180))
+    screen.blit(info, (10, 10))
+
+    stats = small_font.render(f"Warnings: {total_warnings}", True, (150, 150, 180))
+    screen.blit(stats, (10, 40))
+
+    pygame.display.update()
+    clock.tick(30)
+
+cap.release()
+pygame.quit()
+sys.exit()

projects/2026_AI營1/劉宗穎/脸部侦测/face_mesh_connections.py

@@ -0,0 +1,500 @@
+# Copyright 2021 The MediaPipe Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""MediaPipe FaceMesh connections."""
+
+FACEMESH_LIPS = frozenset([(61, 146), (146, 91), (91, 181), (181, 84), (84, 17),
+                           (17, 314), (314, 405), (405, 321), (321, 375),
+                           (375, 291), (61, 185), (185, 40), (40, 39), (39, 37),
+                           (37, 0), (0, 267),
+                           (267, 269), (269, 270), (270, 409), (409, 291),
+                           (78, 95), (95, 88), (88, 178), (178, 87), (87, 14),
+                           (14, 317), (317, 402), (402, 318), (318, 324),
+                           (324, 308), (78, 191), (191, 80), (80, 81), (81, 82),
+                           (82, 13), (13, 312), (312, 311), (311, 310),
+                           (310, 415), (415, 308)])
+
+FACEMESH_LEFT_EYE = frozenset([(263, 249), (249, 390), (390, 373), (373, 374),
+                               (374, 380), (380, 381), (381, 382), (382, 362),
+                               (263, 466), (466, 388), (388, 387), (387, 386),
+                               (386, 385), (385, 384), (384, 398), (398, 362)])
+
+FACEMESH_LEFT_IRIS = frozenset([(474, 475), (475, 476), (476, 477),
+                                 (477, 474)])
+
+FACEMESH_LEFT_EYEBROW = frozenset([(276, 283), (283, 282), (282, 295),
+                                   (295, 285), (300, 293), (293, 334),
+                                   (334, 296), (296, 336)])
+
+FACEMESH_RIGHT_EYE = frozenset([(33, 7), (7, 163), (163, 144), (144, 145),
+                                (145, 153), (153, 154), (154, 155), (155, 133),
+                                (33, 246), (246, 161), (161, 160), (160, 159),
+                                (159, 158), (158, 157), (157, 173), (173, 133)])
+
+FACEMESH_RIGHT_EYEBROW = frozenset([(46, 53), (53, 52), (52, 65), (65, 55),
+                                    (70, 63), (63, 105), (105, 66), (66, 107)])
+
+FACEMESH_RIGHT_IRIS = frozenset([(469, 470), (470, 471), (471, 472),
+                                 (472, 469)])
+
+FACEMESH_FACE_OVAL = frozenset([(10, 338), (338, 297), (297, 332), (332, 284),
+                                (284, 251), (251, 389), (389, 356), (356, 454),
+                                (454, 323), (323, 361), (361, 288), (288, 397),
+                                (397, 365), (365, 379), (379, 378), (378, 400),
+                                (400, 377), (377, 152), (152, 148), (148, 176),
+                                (176, 149), (149, 150), (150, 136), (136, 172),
+                                (172, 58), (58, 132), (132, 93), (93, 234),
+                                (234, 127), (127, 162), (162, 21), (21, 54),
+                                (54, 103), (103, 67), (67, 109), (109, 10)])
+
+FACEMESH_NOSE = frozenset([(168, 6), (6, 197), (197, 195), (195, 5),
+                           (5, 4), (4, 1), (1, 19), (19, 94), (94, 2), (98, 97),
+                           (97, 2), (2, 326), (326, 327), (327, 294),
+                           (294, 278), (278, 344), (344, 440), (440, 275),
+                           (275, 4), (4, 45), (45, 220), (220, 115), (115, 48),
+                           (48, 64), (64, 98)])
+
+FACEMESH_CONTOURS = frozenset().union(*[
+    FACEMESH_LIPS, FACEMESH_LEFT_EYE, FACEMESH_LEFT_EYEBROW, FACEMESH_RIGHT_EYE,
+    FACEMESH_RIGHT_EYEBROW, FACEMESH_FACE_OVAL
+])
+
+FACEMESH_IRISES = frozenset().union(*[FACEMESH_LEFT_IRIS, FACEMESH_RIGHT_IRIS])
+
+FACEMESH_TESSELATION = frozenset([
+    (127, 34),  (34, 139),  (139, 127), (11, 0),    (0, 37),    (37, 11),
+    (232, 231), (231, 120), (120, 232), (72, 37),   (37, 39),   (39, 72),
+    (128, 121), (121, 47),  (47, 128),  (232, 121), (121, 128), (128, 232),
+    (104, 69),  (69, 67),   (67, 104),  (175, 171), (171, 148), (148, 175),
+    (118, 50),  (50, 101),  (101, 118), (73, 39),   (39, 40),   (40, 73),
+    (9, 151),   (151, 108), (108, 9),   (48, 115),  (115, 131), (131, 48),
+    (194, 204), (204, 211), (211, 194), (74, 40),   (40, 185),  (185, 74),
+    (80, 42),   (42, 183),  (183, 80),  (40, 92),   (92, 186),  (186, 40),
+    (230, 229), (229, 118), (118, 230), (202, 212), (212, 214), (214, 202),
+    (83, 18),   (18, 17),   (17, 83),   (76, 61),   (61, 146),  (146, 76),
+    (160, 29),  (29, 30),   (30, 160),  (56, 157),  (157, 173), (173, 56),
+    (106, 204), (204, 194), (194, 106), (135, 214), (214, 192), (192, 135),
+    (203, 165), (165, 98),  (98, 203),  (21, 71),   (71, 68),   (68, 21),
+    (51, 45),   (45, 4),    (4, 51),    (144, 24),  (24, 23),   (23, 144),
+    (77, 146),  (146, 91),  (91, 77),   (205, 50),  (50, 187),  (187, 205),
+    (201, 200), (200, 18),  (18, 201),  (91, 106),  (106, 182), (182, 91),
+    (90, 91),   (91, 181),  (181, 90),  (85, 84),   (84, 17),   (17, 85),
+    (206, 203), (203, 36),  (36, 206),  (148, 171), (171, 140), (140, 148),
+    (92, 40),   (40, 39),   (39, 92),   (193, 189), (189, 244), (244, 193),
+    (159, 158), (158, 28),  (28, 159),  (247, 246), (246, 161), (161, 247),
+    (236, 3),   (3, 196),   (196, 236), (54, 68),   (68, 104),  (104, 54),
+    (193, 168), (168, 8),   (8, 193),   (117, 228), (228, 31),  (31, 117),
+    (189, 193), (193, 55),  (55, 189),  (98, 97),   (97, 99),   (99, 98),
+    (126, 47),  (47, 100),  (100, 126), (166, 79),  (79, 218),  (218, 166),
+    (155, 154), (154, 26),  (26, 155),  (209, 49),  (49, 131),  (131, 209),
+    (135, 136), (136, 150), (150, 135), (47, 126),  (126, 217), (217, 47),
+    (223, 52),  (52, 53),   (53, 223),  (45, 51),   (51, 134),  (134, 45),
+    (211, 170), (170, 140), (140, 211), (67, 69),   (69, 108),  (108, 67),
+    (43, 106),  (106, 91),  (91, 43),   (230, 119), (119, 120), (120, 230),
+    (226, 130), (130, 247), (247, 226), (63, 53),   (53, 52),   (52, 63),
+    (238, 20),  (20, 242),  (242, 238), (46, 70),   (70, 156),  (156, 46),
+    (78, 62),   (62, 96),   (96, 78),   (46, 53),   (53, 63),   (63, 46),
+    (143, 34),  (34, 227),  (227, 143), (123, 117), (117, 111), (111, 123),
+    (44, 125),  (125, 19),  (19, 44),   (236, 134), (134, 51),  (51, 236),
+    (216, 206), (206, 205), (205, 216), (154, 153), (153, 22),  (22, 154),
+    (39, 37),   (37, 167),  (167, 39),  (200, 201), (201, 208), (208, 200),
+    (36, 142),  (142, 100), (100, 36),  (57, 212),  (212, 202), (202, 57),
+    (20, 60),   (60, 99),   (99, 20),   (28, 158),  (158, 157), (157, 28),
+    (35, 226),  (226, 113), (113, 35),  (160, 159), (159, 27),  (27, 160),
+    (204, 202), (202, 210), (210, 204), (113, 225), (225, 46),  (46, 113),
+    (43, 202),  (202, 204), (204, 43),  (62, 76),   (76, 77),   (77, 62),
+    (137, 123), (123, 116), (116, 137), (41, 38),   (38, 72),   (72, 41),
+    (203, 129), (129, 142), (142, 203), (64, 98),   (98, 240),  (240, 64),
+    (49, 102),  (102, 64),  (64, 49),   (41, 73),   (73, 74),   (74, 41),
+    (212, 216), (216, 207), (207, 212), (42, 74),   (74, 184),  (184, 42),
+    (169, 170), (170, 211), (211, 169), (170, 149), (149, 176), (176, 170),
+    (105, 66),  (66, 69),   (69, 105),  (122, 6),   (6, 168),   (168, 122),
+    (123, 147), (147, 187), (187, 123), (96, 77),   (77, 90),   (90, 96),
+    (65, 55),   (55, 107),  (107, 65),  (89, 90),   (90, 180),  (180, 89),
+    (101, 100), (100, 120), (120, 101), (63, 105),  (105, 104), (104, 63),
+    (93, 137),  (137, 227), (227, 93),  (15, 86),   (86, 85),   (85, 15),
+    (129, 102), (102, 49),  (49, 129),  (14, 87),   (87, 86),   (86, 14),
+    (55, 8),    (8, 9),     (9, 55),    (100, 47),  (47, 121),  (121, 100),
+    (145, 23),  (23, 22),   (22, 145),  (88, 89),   (89, 179),  (179, 88),
+    (6, 122),   (122, 196), (196, 6),   (88, 95),   (95, 96),   (96, 88),
+    (138, 172), (172, 136), (136, 138), (215, 58),  (58, 172),  (172, 215),
+    (115, 48),  (48, 219),  (219, 115), (42, 80),   (80, 81),   (81, 42),
+    (195, 3),   (3, 51),    (51, 195),  (43, 146),  (146, 61),  (61, 43),
+    (171, 175), (175, 199), (199, 171), (81, 82),   (82, 38),   (38, 81),
+    (53, 46),   (46, 225),  (225, 53),  (144, 163), (163, 110), (110, 144),
+    (52, 65),   (65, 66),   (66, 52),   (229, 228), (228, 117), (117, 229),
+    (34, 127),  (127, 234), (234, 34),  (107, 108), (108, 69),  (69, 107),
+    (109, 108), (108, 151), (151, 109), (48, 64),   (64, 235),  (235, 48),
+    (62, 78),   (78, 191),  (191, 62),  (129, 209), (209, 126), (126, 129),
+    (111, 35),  (35, 143),  (143, 111), (117, 123), (123, 50),  (50, 117),
+    (222, 65),  (65, 52),   (52, 222),  (19, 125),  (125, 141), (141, 19),
+    (221, 55),  (55, 65),   (65, 221),  (3, 195),   (195, 197), (197, 3),
+    (25, 7),    (7, 33),    (33, 25),   (220, 237), (237, 44),  (44, 220),
+    (70, 71),   (71, 139),  (139, 70),  (122, 193), (193, 245), (245, 122),
+    (247, 130), (130, 33),  (33, 247),  (71, 21),   (21, 162),  (162, 71),
+    (170, 169), (169, 150), (150, 170), (188, 174), (174, 196), (196, 188),
+    (216, 186), (186, 92),  (92, 216),  (2, 97),    (97, 167),  (167, 2),
+    (141, 125), (125, 241), (241, 141), (164, 167), (167, 37),  (37, 164),
+    (72, 38),   (38, 12),   (12, 72),   (38, 82),   (82, 13),   (13, 38),
+    (63, 68),   (68, 71),   (71, 63),   (226, 35),  (35, 111),  (111, 226),
+    (101, 50),  (50, 205),  (205, 101), (206, 92),  (92, 165),  (165, 206),
+    (209, 198), (198, 217), (217, 209), (165, 167), (167, 97),  (97, 165),
+    (220, 115), (115, 218), (218, 220), (133, 112), (112, 243), (243, 133),
+    (239, 238), (238, 241), (241, 239), (214, 135), (135, 169), (169, 214),
+    (190, 173), (173, 133), (133, 190), (171, 208), (208, 32),  (32, 171),
+    (125, 44),  (44, 237),  (237, 125), (86, 87),   (87, 178),  (178, 86),
+    (85, 86),   (86, 179),  (179, 85),  (84, 85),   (85, 180),  (180, 84),
+    (83, 84),   (84, 181),  (181, 83),  (201, 83),  (83, 182),  (182, 201),
+    (137, 93),  (93, 132),  (132, 137), (76, 62),   (62, 183),  (183, 76),
+    (61, 76),   (76, 184),  (184, 61),  (57, 61),   (61, 185),  (185, 57),
+    (212, 57),  (57, 186),  (186, 212), (214, 207), (207, 187), (187, 214),
+    (34, 143),  (143, 156), (156, 34),  (79, 239),  (239, 237), (237, 79),
+    (123, 137), (137, 177), (177, 123), (44, 1),    (1, 4),     (4, 44),
+    (201, 194), (194, 32),  (32, 201),  (64, 102),  (102, 129), (129, 64),
+    (213, 215), (215, 138), (138, 213), (59, 166),  (166, 219), (219, 59),
+    (242, 99),  (99, 97),   (97, 242),  (2, 94),    (94, 141),  (141, 2),
+    (75, 59),   (59, 235),  (235, 75),  (24, 110),  (110, 228), (228, 24),
+    (25, 130),  (130, 226), (226, 25),  (23, 24),   (24, 229),  (229, 23),
+    (22, 23),   (23, 230),  (230, 22),  (26, 22),   (22, 231),  (231, 26),
+    (112, 26),  (26, 232),  (232, 112), (189, 190), (190, 243), (243, 189),
+    (221, 56),  (56, 190),  (190, 221), (28, 56),   (56, 221),  (221, 28),
+    (27, 28),   (28, 222),  (222, 27),  (29, 27),   (27, 223),  (223, 29),
+    (30, 29),   (29, 224),  (224, 30),  (247, 30),  (30, 225),  (225, 247),
+    (238, 79),  (79, 20),   (20, 238),  (166, 59),  (59, 75),   (75, 166),
+    (60, 75),   (75, 240),  (240, 60),  (147, 177), (177, 215), (215, 147),
+    (20, 79),   (79, 166),  (166, 20),  (187, 147), (147, 213), (213, 187),
+    (112, 233), (233, 244), (244, 112), (233, 128), (128, 245), (245, 233),
+    (128, 114), (114, 188), (188, 128), (114, 217), (217, 174), (174, 114),
+    (131, 115), (115, 220), (220, 131), (217, 198), (198, 236), (236, 217),
+    (198, 131), (131, 134), (134, 198), (177, 132), (132, 58),  (58, 177),
+    (143, 35),  (35, 124),  (124, 143), (110, 163), (163, 7),   (7, 110),
+    (228, 110), (110, 25),  (25, 228),  (356, 389), (389, 368), (368, 356),
+    (11, 302),  (302, 267), (267, 11),  (452, 350), (350, 349), (349, 452),
+    (302, 303), (303, 269), (269, 302), (357, 343), (343, 277), (277, 357),
+    (452, 453), (453, 357), (357, 452), (333, 332), (332, 297), (297, 333),
+    (175, 152), (152, 377), (377, 175), (347, 348), (348, 330), (330, 347),
+    (303, 304), (304, 270), (270, 303), (9, 336),   (336, 337), (337, 9),
+    (278, 279), (279, 360), (360, 278), (418, 262), (262, 431), (431, 418),
+    (304, 408), (408, 409), (409, 304), (310, 415), (415, 407), (407, 310),
+    (270, 409), (409, 410), (410, 270), (450, 348), (348, 347), (347, 450),
+    (422, 430), (430, 434), (434, 422), (313, 314), (314, 17),  (17, 313),
+    (306, 307), (307, 375), (375, 306), (387, 388), (388, 260), (260, 387),
+    (286, 414), (414, 398), (398, 286), (335, 406), (406, 418), (418, 335),
+    (364, 367), (367, 416), (416, 364), (423, 358), (358, 327), (327, 423),
+    (251, 284), (284, 298), (298, 251), (281, 5),   (5, 4),     (4, 281),
+    (373, 374), (374, 253), (253, 373), (307, 320), (320, 321), (321, 307),
+    (425, 427), (427, 411), (411, 425), (421, 313), (313, 18),  (18, 421),
+    (321, 405), (405, 406), (406, 321), (320, 404), (404, 405), (405, 320),
+    (315, 16),  (16, 17),   (17, 315),  (426, 425), (425, 266), (266, 426),
+    (377, 400), (400, 369), (369, 377), (322, 391), (391, 269), (269, 322),
+    (417, 465), (465, 464), (464, 417), (386, 257), (257, 258), (258, 386),
+    (466, 260), (260, 388), (388, 466), (456, 399), (399, 419), (419, 456),
+    (284, 332), (332, 333), (333, 284), (417, 285), (285, 8),   (8, 417),
+    (346, 340), (340, 261), (261, 346), (413, 441), (441, 285), (285, 413),
+    (327, 460), (460, 328), (328, 327), (355, 371), (371, 329), (329, 355),
+    (392, 439), (439, 438), (438, 392), (382, 341), (341, 256), (256, 382),
+    (429, 420), (420, 360), (360, 429), (364, 394), (394, 379), (379, 364),
+    (277, 343), (343, 437), (437, 277), (443, 444), (444, 283), (283, 443),
+    (275, 440), (440, 363), (363, 275), (431, 262), (262, 369), (369, 431),
+    (297, 338), (338, 337), (337, 297), (273, 375), (375, 321), (321, 273),
+    (450, 451), (451, 349), (349, 450), (446, 342), (342, 467), (467, 446),
+    (293, 334), (334, 282), (282, 293), (458, 461), (461, 462), (462, 458),
+    (276, 353), (353, 383), (383, 276), (308, 324), (324, 325), (325, 308),
+    (276, 300), (300, 293), (293, 276), (372, 345), (345, 447), (447, 372),
+    (352, 345), (345, 340), (340, 352), (274, 1),   (1, 19),    (19, 274),
+    (456, 248), (248, 281), (281, 456), (436, 427), (427, 425), (425, 436),
+    (381, 256), (256, 252), (252, 381), (269, 391), (391, 393), (393, 269),
+    (200, 199), (199, 428), (428, 200), (266, 330), (330, 329), (329, 266),
+    (287, 273), (273, 422), (422, 287), (250, 462), (462, 328), (328, 250),
+    (258, 286), (286, 384), (384, 258), (265, 353), (353, 342), (342, 265),
+    (387, 259), (259, 257), (257, 387), (424, 431), (431, 430), (430, 424),
+    (342, 353), (353, 276), (276, 342), (273, 335), (335, 424), (424, 273),
+    (292, 325), (325, 307), (307, 292), (366, 447), (447, 345), (345, 366),
+    (271, 303), (303, 302), (302, 271), (423, 266), (266, 371), (371, 423),
+    (294, 455), (455, 460), (460, 294), (279, 278), (278, 294), (294, 279),
+    (271, 272), (272, 304), (304, 271), (432, 434), (434, 427), (427, 432),
+    (272, 407), (407, 408), (408, 272), (394, 430), (430, 431), (431, 394),
+    (395, 369), (369, 400), (400, 395), (334, 333), (333, 299), (299, 334),
+    (351, 417), (417, 168), (168, 351), (352, 280), (280, 411), (411, 352),
+    (325, 319), (319, 320), (320, 325), (295, 296), (296, 336), (336, 295),
+    (319, 403), (403, 404), (404, 319), (330, 348), (348, 349), (349, 330),
+    (293, 298), (298, 333), (333, 293), (323, 454), (454, 447), (447, 323),
+    (15, 16),   (16, 315),  (315, 15),  (358, 429), (429, 279), (279, 358),
+    (14, 15),   (15, 316),  (316, 14),  (285, 336), (336, 9),   (9, 285),
+    (329, 349), (349, 350), (350, 329), (374, 380), (380, 252), (252, 374),
+    (318, 402), (402, 403), (403, 318), (6, 197),   (197, 419), (419, 6),
+    (318, 319), (319, 325), (325, 318), (367, 364), (364, 365), (365, 367),
+    (435, 367), (367, 397), (397, 435), (344, 438), (438, 439), (439, 344),
+    (272, 271), (271, 311), (311, 272), (195, 5),   (5, 281),   (281, 195),
+    (273, 287), (287, 291), (291, 273), (396, 428), (428, 199), (199, 396),
+    (311, 271), (271, 268), (268, 311), (283, 444), (444, 445), (445, 283),
+    (373, 254), (254, 339), (339, 373), (282, 334), (334, 296), (296, 282),
+    (449, 347), (347, 346), (346, 449), (264, 447), (447, 454), (454, 264),
+    (336, 296), (296, 299), (299, 336), (338, 10),  (10, 151),  (151, 338),
+    (278, 439), (439, 455), (455, 278), (292, 407), (407, 415), (415, 292),
+    (358, 371), (371, 355), (355, 358), (340, 345), (345, 372), (372, 340),
+    (346, 347), (347, 280), (280, 346), (442, 443), (443, 282), (282, 442),
+    (19, 94),   (94, 370),  (370, 19),  (441, 442), (442, 295), (295, 441),
+    (248, 419), (419, 197), (197, 248), (263, 255), (255, 359), (359, 263),
+    (440, 275), (275, 274), (274, 440), (300, 383), (383, 368), (368, 300),
+    (351, 412), (412, 465), (465, 351), (263, 467), (467, 466), (466, 263),
+    (301, 368), (368, 389), (389, 301), (395, 378), (378, 379), (379, 395),
+    (412, 351), (351, 419), (419, 412), (436, 426), (426, 322), (322, 436),
+    (2, 164),   (164, 393), (393, 2),   (370, 462), (462, 461), (461, 370),
+    (164, 0),   (0, 267),   (267, 164), (302, 11),  (11, 12),   (12, 302),
+    (268, 12),  (12, 13),   (13, 268),  (293, 300), (300, 301), (301, 293),
+    (446, 261), (261, 340), (340, 446), (330, 266), (266, 425), (425, 330),
+    (426, 423), (423, 391), (391, 426), (429, 355), (355, 437), (437, 429),
+    (391, 327), (327, 326), (326, 391), (440, 457), (457, 438), (438, 440),
+    (341, 382), (382, 362), (362, 341), (459, 457), (457, 461), (461, 459),
+    (434, 430), (430, 394), (394, 434), (414, 463), (463, 362), (362, 414),
+    (396, 369), (369, 262), (262, 396), (354, 461), (461, 457), (457, 354),
+    (316, 403), (403, 402), (402, 316), (315, 404), (404, 403), (403, 315),
+    (314, 405), (405, 404), (404, 314), (313, 406), (406, 405), (405, 313),
+    (421, 418), (418, 406), (406, 421), (366, 401), (401, 361), (361, 366),
+    (306, 408), (408, 407), (407, 306), (291, 409), (409, 408), (408, 291),
+    (287, 410), (410, 409), (409, 287), (432, 436), (436, 410), (410, 432),
+    (434, 416), (416, 411), (411, 434), (264, 368), (368, 383), (383, 264),
+    (309, 438), (438, 457), (457, 309), (352, 376), (376, 401), (401, 352),
+    (274, 275), (275, 4),   (4, 274),   (421, 428), (428, 262), (262, 421),
+    (294, 327), (327, 358), (358, 294), (433, 416), (416, 367), (367, 433),
+    (289, 455), (455, 439), (439, 289), (462, 370), (370, 326), (326, 462),
+    (2, 326),   (326, 370), (370, 2),   (305, 460), (460, 455), (455, 305),
+    (254, 449), (449, 448), (448, 254), (255, 261), (261, 446), (446, 255),
+    (253, 450), (450, 449), (449, 253), (252, 451), (451, 450), (450, 252),
+    (256, 452), (452, 451), (451, 256), (341, 453), (453, 452), (452, 341),
+    (413, 464), (464, 463), (463, 413), (441, 413), (413, 414), (414, 441),
+    (258, 442), (442, 441), (441, 258), (257, 443), (443, 442), (442, 257),
+    (259, 444), (444, 443), (443, 259), (260, 445), (445, 444), (444, 260),
+    (467, 342), (342, 445), (445, 467), (459, 458), (458, 250), (250, 459),
+    (289, 392), (392, 290), (290, 289), (290, 328), (328, 460), (460, 290),
+    (376, 433), (433, 435), (435, 376), (250, 290), (290, 392), (392, 250),
+    (411, 416), (416, 433), (433, 411), (341, 463), (463, 464), (464, 341),
+    (453, 464), (464, 465), (465, 453), (357, 465), (465, 412), (412, 357),
+    (343, 412), (412, 399), (399, 343), (360, 363), (363, 440), (440, 360),
+    (437, 399), (399, 456), (456, 437), (420, 456), (456, 363), (363, 420),
+    (401, 435), (435, 288), (288, 401), (372, 383), (383, 353), (353, 372),
+    (339, 255), (255, 249), (249, 339), (448, 261), (261, 255), (255, 448),
+    (133, 243), (243, 190), (190, 133), (133, 155), (155, 112), (112, 133),
+    (33, 246),  (246, 247), (247, 33),  (33, 130),  (130, 25),  (25, 33),
+    (398, 384), (384, 286), (286, 398), (362, 398), (398, 414), (414, 362),
+    (362, 463), (463, 341), (341, 362), (263, 359), (359, 467), (467, 263),
+    (263, 249), (249, 255), (255, 263), (466, 467), (467, 260), (260, 466),
+    (75, 60),   (60, 166),  (166, 75),  (238, 239), (239, 79),  (79, 238),
+    (162, 127), (127, 139), (139, 162), (72, 11),   (11, 37),   (37, 72),
+    (121, 232), (232, 120), (120, 121), (73, 72),   (72, 39),   (39, 73),
+    (114, 128), (128, 47),  (47, 114),  (233, 232), (232, 128), (128, 233),
+    (103, 104), (104, 67),  (67, 103),  (152, 175), (175, 148), (148, 152),
+    (119, 118), (118, 101), (101, 119), (74, 73),   (73, 40),   (40, 74),
+    (107, 9),   (9, 108),   (108, 107), (49, 48),   (48, 131),  (131, 49),
+    (32, 194),  (194, 211), (211, 32),  (184, 74),  (74, 185),  (185, 184),
+    (191, 80),  (80, 183),  (183, 191), (185, 40),  (40, 186),  (186, 185),
+    (119, 230), (230, 118), (118, 119), (210, 202), (202, 214), (214, 210),
+    (84, 83),   (83, 17),   (17, 84),   (77, 76),   (76, 146),  (146, 77),
+    (161, 160), (160, 30),  (30, 161),  (190, 56),  (56, 173),  (173, 190),
+    (182, 106), (106, 194), (194, 182), (138, 135), (135, 192), (192, 138),
+    (129, 203), (203, 98),  (98, 129),  (54, 21),   (21, 68),   (68, 54),
+    (5, 51),    (51, 4),    (4, 5),     (145, 144), (144, 23),  (23, 145),
+    (90, 77),   (77, 91),   (91, 90),   (207, 205), (205, 187), (187, 207),
+    (83, 201),  (201, 18),  (18, 83),   (181, 91),  (91, 182),  (182, 181),
+    (180, 90),  (90, 181),  (181, 180), (16, 85),   (85, 17),   (17, 16),
+    (205, 206), (206, 36),  (36, 205),  (176, 148), (148, 140), (140, 176),
+    (165, 92),  (92, 39),   (39, 165),  (245, 193), (193, 244), (244, 245),
+    (27, 159),  (159, 28),  (28, 27),   (30, 247),  (247, 161), (161, 30),
+    (174, 236), (236, 196), (196, 174), (103, 54),  (54, 104),  (104, 103),
+    (55, 193),  (193, 8),   (8, 55),    (111, 117), (117, 31),  (31, 111),
+    (221, 189), (189, 55),  (55, 221),  (240, 98),  (98, 99),   (99, 240),
+    (142, 126), (126, 100), (100, 142), (219, 166), (166, 218), (218, 219),
+    (112, 155), (155, 26),  (26, 112),  (198, 209), (209, 131), (131, 198),
+    (169, 135), (135, 150), (150, 169), (114, 47),  (47, 217),  (217, 114),
+    (224, 223), (223, 53),  (53, 224),  (220, 45),  (45, 134),  (134, 220),
+    (32, 211),  (211, 140), (140, 32),  (109, 67),  (67, 108),  (108, 109),
+    (146, 43),  (43, 91),   (91, 146),  (231, 230), (230, 120), (120, 231),
+    (113, 226), (226, 247), (247, 113), (105, 63),  (63, 52),   (52, 105),
+    (241, 238), (238, 242), (242, 241), (124, 46),  (46, 156),  (156, 124),
+    (95, 78),   (78, 96),   (96, 95),   (70, 46),   (46, 63),   (63, 70),
+    (116, 143), (143, 227), (227, 116), (116, 123), (123, 111), (111, 116),
+    (1, 44),    (44, 19),   (19, 1),    (3, 236),   (236, 51),  (51, 3),
+    (207, 216), (216, 205), (205, 207), (26, 154),  (154, 22),  (22, 26),
+    (165, 39),  (39, 167),  (167, 165), (199, 200), (200, 208), (208, 199),
+    (101, 36),  (36, 100),  (100, 101), (43, 57),   (57, 202),  (202, 43),
+    (242, 20),  (20, 99),   (99, 242),  (56, 28),   (28, 157),  (157, 56),
+    (124, 35),  (35, 113),  (113, 124), (29, 160),  (160, 27),  (27, 29),
+    (211, 204), (204, 210), (210, 211), (124, 113), (113, 46),  (46, 124),
+    (106, 43),  (43, 204),  (204, 106), (96, 62),   (62, 77),   (77, 96),
+    (227, 137), (137, 116), (116, 227), (73, 41),   (41, 72),   (72, 73),
+    (36, 203),  (203, 142), (142, 36),  (235, 64),  (64, 240),  (240, 235),
+    (48, 49),   (49, 64),   (64, 48),   (42, 41),   (41, 74),   (74, 42),
+    (214, 212), (212, 207), (207, 214), (183, 42),  (42, 184),  (184, 183),
+    (210, 169), (169, 211), (211, 210), (140, 170), (170, 176), (176, 140),
+    (104, 105), (105, 69),  (69, 104),  (193, 122), (122, 168), (168, 193),
+    (50, 123),  (123, 187), (187, 50),  (89, 96),   (96, 90),   (90, 89),
+    (66, 65),   (65, 107),  (107, 66),  (179, 89),  (89, 180),  (180, 179),
+    (119, 101), (101, 120), (120, 119), (68, 63),   (63, 104),  (104, 68),
+    (234, 93),  (93, 227),  (227, 234), (16, 15),   (15, 85),   (85, 16),
+    (209, 129), (129, 49),  (49, 209),  (15, 14),   (14, 86),   (86, 15),
+    (107, 55),  (55, 9),    (9, 107),   (120, 100), (100, 121), (121, 120),
+    (153, 145), (145, 22),  (22, 153),  (178, 88),  (88, 179),  (179, 178),
+    (197, 6),   (6, 196),   (196, 197), (89, 88),   (88, 96),   (96, 89),
+    (135, 138), (138, 136), (136, 135), (138, 215), (215, 172), (172, 138),
+    (218, 115), (115, 219), (219, 218), (41, 42),   (42, 81),   (81, 41),
+    (5, 195),   (195, 51),  (51, 5),    (57, 43),   (43, 61),   (61, 57),
+    (208, 171), (171, 199), (199, 208), (41, 81),   (81, 38),   (38, 41),
+    (224, 53),  (53, 225),  (225, 224), (24, 144),  (144, 110), (110, 24),
+    (105, 52),  (52, 66),   (66, 105),  (118, 229), (229, 117), (117, 118),
+    (227, 34),  (34, 234),  (234, 227), (66, 107),  (107, 69),  (69, 66),
+    (10, 109),  (109, 151), (151, 10),  (219, 48),  (48, 235),  (235, 219),
+    (183, 62),  (62, 191),  (191, 183), (142, 129), (129, 126), (126, 142),
+    (116, 111), (111, 143), (143, 116), (118, 117), (117, 50),  (50, 118),
+    (223, 222), (222, 52),  (52, 223),  (94, 19),   (19, 141),  (141, 94),
+    (222, 221), (221, 65),  (65, 222),  (196, 3),   (3, 197),   (197, 196),
+    (45, 220),  (220, 44),  (44, 45),   (156, 70),  (70, 139),  (139, 156),
+    (188, 122), (122, 245), (245, 188), (139, 71),  (71, 162),  (162, 139),
+    (149, 170), (170, 150), (150, 149), (122, 188), (188, 196), (196, 122),
+    (206, 216), (216, 92),  (92, 206),  (164, 2),   (2, 167),   (167, 164),
+    (242, 141), (141, 241), (241, 242), (0, 164),   (164, 37),  (37, 0),
+    (11, 72),   (72, 12),   (12, 11),   (12, 38),   (38, 13),   (13, 12),
+    (70, 63),   (63, 71),   (71, 70),   (31, 226),  (226, 111), (111, 31),
+    (36, 101),  (101, 205), (205, 36),  (203, 206), (206, 165), (165, 203),
+    (126, 209), (209, 217), (217, 126), (98, 165),  (165, 97),  (97, 98),
+    (237, 220), (220, 218), (218, 237), (237, 239), (239, 241), (241, 237),
+    (210, 214), (214, 169), (169, 210), (140, 171), (171, 32),  (32, 140),
+    (241, 125), (125, 237), (237, 241), (179, 86),  (86, 178),  (178, 179),
+    (180, 85),  (85, 179),  (179, 180), (181, 84),  (84, 180),  (180, 181),
+    (182, 83),  (83, 181),  (181, 182), (194, 201), (201, 182), (182, 194),
+    (177, 137), (137, 132), (132, 177), (184, 76),  (76, 183),  (183, 184),
+    (185, 61),  (61, 184),  (184, 185), (186, 57),  (57, 185),  (185, 186),
+    (216, 212), (212, 186), (186, 216), (192, 214), (214, 187), (187, 192),
+    (139, 34),  (34, 156),  (156, 139), (218, 79),  (79, 237),  (237, 218),
+    (147, 123), (123, 177), (177, 147), (45, 44),   (44, 4),    (4, 45),
+    (208, 201), (201, 32),  (32, 208),  (98, 64),   (64, 129),  (129, 98),
+    (192, 213), (213, 138), (138, 192), (235, 59),  (59, 219),  (219, 235),
+    (141, 242), (242, 97),  (97, 141),  (97, 2),    (2, 141),   (141, 97),
+    (240, 75),  (75, 235),  (235, 240), (229, 24),  (24, 228),  (228, 229),
+    (31, 25),   (25, 226),  (226, 31),  (230, 23),  (23, 229),  (229, 230),
+    (231, 22),  (22, 230),  (230, 231), (232, 26),  (26, 231),  (231, 232),
+    (233, 112), (112, 232), (232, 233), (244, 189), (189, 243), (243, 244),
+    (189, 221), (221, 190), (190, 189), (222, 28),  (28, 221),  (221, 222),
+    (223, 27),  (27, 222),  (222, 223), (224, 29),  (29, 223),  (223, 224),
+    (225, 30),  (30, 224),  (224, 225), (113, 247), (247, 225), (225, 113),
+    (99, 60),   (60, 240),  (240, 99),  (213, 147), (147, 215), (215, 213),
+    (60, 20),   (20, 166),  (166, 60),  (192, 187), (187, 213), (213, 192),
+    (243, 112), (112, 244), (244, 243), (244, 233), (233, 245), (245, 244),
+    (245, 128), (128, 188), (188, 245), (188, 114), (114, 174), (174, 188),
+    (134, 131), (131, 220), (220, 134), (174, 217), (217, 236), (236, 174),
+    (236, 198), (198, 134), (134, 236), (215, 177), (177, 58),  (58, 215),
+    (156, 143), (143, 124), (124, 156), (25, 110),  (110, 7),   (7, 25),
+    (31, 228),  (228, 25),  (25, 31),   (264, 356), (356, 368), (368, 264),
+    (0, 11),    (11, 267),  (267, 0),   (451, 452), (452, 349), (349, 451),
+    (267, 302), (302, 269), (269, 267), (350, 357), (357, 277), (277, 350),
+    (350, 452), (452, 357), (357, 350), (299, 333), (333, 297), (297, 299),
+    (396, 175), (175, 377), (377, 396), (280, 347), (347, 330), (330, 280),
+    (269, 303), (303, 270), (270, 269), (151, 9),   (9, 337),   (337, 151),
+    (344, 278), (278, 360), (360, 344), (424, 418), (418, 431), (431, 424),
+    (270, 304), (304, 409), (409, 270), (272, 310), (310, 407), (407, 272),
+    (322, 270), (270, 410), (410, 322), (449, 450), (450, 347), (347, 449),
+    (432, 422), (422, 434), (434, 432), (18, 313),  (313, 17),  (17, 18),
+    (291, 306), (306, 375), (375, 291), (259, 387), (387, 260), (260, 259),
+    (424, 335), (335, 418), (418, 424), (434, 364), (364, 416), (416, 434),
+    (391, 423), (423, 327), (327, 391), (301, 251), (251, 298), (298, 301),
+    (275, 281), (281, 4),   (4, 275),   (254, 373), (373, 253), (253, 254),
+    (375, 307), (307, 321), (321, 375), (280, 425), (425, 411), (411, 280),
+    (200, 421), (421, 18),  (18, 200),  (335, 321), (321, 406), (406, 335),
+    (321, 320), (320, 405), (405, 321), (314, 315), (315, 17),  (17, 314),
+    (423, 426), (426, 266), (266, 423), (396, 377), (377, 369), (369, 396),
+    (270, 322), (322, 269), (269, 270), (413, 417), (417, 464), (464, 413),
+    (385, 386), (386, 258), (258, 385), (248, 456), (456, 419), (419, 248),
+    (298, 284), (284, 333), (333, 298), (168, 417), (417, 8),   (8, 168),
+    (448, 346), (346, 261), (261, 448), (417, 413), (413, 285), (285, 417),
+    (326, 327), (327, 328), (328, 326), (277, 355), (355, 329), (329, 277),
+    (309, 392), (392, 438), (438, 309), (381, 382), (382, 256), (256, 381),
+    (279, 429), (429, 360), (360, 279), (365, 364), (364, 379), (379, 365),
+    (355, 277), (277, 437), (437, 355), (282, 443), (443, 283), (283, 282),
+    (281, 275), (275, 363), (363, 281), (395, 431), (431, 369), (369, 395),
+    (299, 297), (297, 337), (337, 299), (335, 273), (273, 321), (321, 335),
+    (348, 450), (450, 349), (349, 348), (359, 446), (446, 467), (467, 359),
+    (283, 293), (293, 282), (282, 283), (250, 458), (458, 462), (462, 250),
+    (300, 276), (276, 383), (383, 300), (292, 308), (308, 325), (325, 292),
+    (283, 276), (276, 293), (293, 283), (264, 372), (372, 447), (447, 264),
+    (346, 352), (352, 340), (340, 346), (354, 274), (274, 19),  (19, 354),
+    (363, 456), (456, 281), (281, 363), (426, 436), (436, 425), (425, 426),
+    (380, 381), (381, 252), (252, 380), (267, 269), (269, 393), (393, 267),
+    (421, 200), (200, 428), (428, 421), (371, 266), (266, 329), (329, 371),
+    (432, 287), (287, 422), (422, 432), (290, 250), (250, 328), (328, 290),
+    (385, 258), (258, 384), (384, 385), (446, 265), (265, 342), (342, 446),
+    (386, 387), (387, 257), (257, 386), (422, 424), (424, 430), (430, 422),
+    (445, 342), (342, 276), (276, 445), (422, 273), (273, 424), (424, 422),
+    (306, 292), (292, 307), (307, 306), (352, 366), (366, 345), (345, 352),
+    (268, 271), (271, 302), (302, 268), (358, 423), (423, 371), (371, 358),
+    (327, 294), (294, 460), (460, 327), (331, 279), (279, 294), (294, 331),
+    (303, 271), (271, 304), (304, 303), (436, 432), (432, 427), (427, 436),
+    (304, 272), (272, 408), (408, 304), (395, 394), (394, 431), (431, 395),
+    (378, 395), (395, 400), (400, 378), (296, 334), (334, 299), (299, 296),
+    (6, 351),   (351, 168), (168, 6),   (376, 352), (352, 411), (411, 376),
+    (307, 325), (325, 320), (320, 307), (285, 295), (295, 336), (336, 285),
+    (320, 319), (319, 404), (404, 320), (329, 330), (330, 349), (349, 329),
+    (334, 293), (293, 333), (333, 334), (366, 323), (323, 447), (447, 366),
+    (316, 15),  (15, 315),  (315, 316), (331, 358), (358, 279), (279, 331),
+    (317, 14),  (14, 316),  (316, 317), (8, 285),   (285, 9),   (9, 8),
+    (277, 329), (329, 350), (350, 277), (253, 374), (374, 252), (252, 253),
+    (319, 318), (318, 403), (403, 319), (351, 6),   (6, 419),   (419, 351),
+    (324, 318), (318, 325), (325, 324), (397, 367), (367, 365), (365, 397),
+    (288, 435), (435, 397), (397, 288), (278, 344), (344, 439), (439, 278),
+    (310, 272), (272, 311), (311, 310), (248, 195), (195, 281), (281, 248),
+    (375, 273), (273, 291), (291, 375), (175, 396), (396, 199), (199, 175),
+    (312, 311), (311, 268), (268, 312), (276, 283), (283, 445), (445, 276),
+    (390, 373), (373, 339), (339, 390), (295, 282), (282, 296), (296, 295),
+    (448, 449), (449, 346), (346, 448), (356, 264), (264, 454), (454, 356),
+    (337, 336), (336, 299), (299, 337), (337, 338), (338, 151), (151, 337),
+    (294, 278), (278, 455), (455, 294), (308, 292), (292, 415), (415, 308),
+    (429, 358), (358, 355), (355, 429), (265, 340), (340, 372), (372, 265),
+    (352, 346), (346, 280), (280, 352), (295, 442), (442, 282), (282, 295),
+    (354, 19),  (19, 370),  (370, 354), (285, 441), (441, 295), (295, 285),
+    (195, 248), (248, 197), (197, 195), (457, 440), (440, 274), (274, 457),
+    (301, 300), (300, 368), (368, 301), (417, 351), (351, 465), (465, 417),
+    (251, 301), (301, 389), (389, 251), (394, 395), (395, 379), (379, 394),
+    (399, 412), (412, 419), (419, 399), (410, 436), (436, 322), (322, 410),
+    (326, 2),   (2, 393),   (393, 326), (354, 370), (370, 461), (461, 354),
+    (393, 164), (164, 267), (267, 393), (268, 302), (302, 12),  (12, 268),
+    (312, 268), (268, 13),  (13, 312),  (298, 293), (293, 301), (301, 298),
+    (265, 446), (446, 340), (340, 265), (280, 330), (330, 425), (425, 280),
+    (322, 426), (426, 391), (391, 322), (420, 429), (429, 437), (437, 420),
+    (393, 391), (391, 326), (326, 393), (344, 440), (440, 438), (438, 344),
+    (458, 459), (459, 461), (461, 458), (364, 434), (434, 394), (394, 364),
+    (428, 396), (396, 262), (262, 428), (274, 354), (354, 457), (457, 274),
+    (317, 316), (316, 402), (402, 317), (316, 315), (315, 403), (403, 316),
+    (315, 314), (314, 404), (404, 315), (314, 313), (313, 405), (405, 314),
+    (313, 421), (421, 406), (406, 313), (323, 366), (366, 361), (361, 323),
+    (292, 306), (306, 407), (407, 292), (306, 291), (291, 408), (408, 306),
+    (291, 287), (287, 409), (409, 291), (287, 432), (432, 410), (410, 287),
+    (427, 434), (434, 411), (411, 427), (372, 264), (264, 383), (383, 372),
+    (459, 309), (309, 457), (457, 459), (366, 352), (352, 401), (401, 366),
+    (1, 274),   (274, 4),   (4, 1),     (418, 421), (421, 262), (262, 418),
+    (331, 294), (294, 358), (358, 331), (435, 433), (433, 367), (367, 435),
+    (392, 289), (289, 439), (439, 392), (328, 462), (462, 326), (326, 328),
+    (94, 2),    (2, 370),   (370, 94),  (289, 305), (305, 455), (455, 289),
+    (339, 254), (254, 448), (448, 339), (359, 255), (255, 446), (446, 359),
+    (254, 253), (253, 449), (449, 254), (253, 252), (252, 450), (450, 253),
+    (252, 256), (256, 451), (451, 252), (256, 341), (341, 452), (452, 256),
+    (414, 413), (413, 463), (463, 414), (286, 441), (441, 414), (414, 286),
+    (286, 258), (258, 441), (441, 286), (258, 257), (257, 442), (442, 258),
+    (257, 259), (259, 443), (443, 257), (259, 260), (260, 444), (444, 259),
+    (260, 467), (467, 445), (445, 260), (309, 459), (459, 250), (250, 309),
+    (305, 289), (289, 290), (290, 305), (305, 290), (290, 460), (460, 305),
+    (401, 376), (376, 435), (435, 401), (309, 250), (250, 392), (392, 309),
+    (376, 411), (411, 433), (433, 376), (453, 341), (341, 464), (464, 453),
+    (357, 453), (453, 465), (465, 357), (343, 357), (357, 412), (412, 343),
+    (437, 343), (343, 399), (399, 437), (344, 360), (360, 440), (440, 344),
+    (420, 437), (437, 456), (456, 420), (360, 420), (420, 363), (363, 360),
+    (361, 401), (401, 288), (288, 361), (265, 372), (372, 353), (353, 265),
+    (390, 339), (339, 249), (249, 390), (339, 448), (448, 255), (255, 339)])

projects/2026_AI營1/劉宗穎/脸部侦测/step1_facemesh.py

@@ -0,0 +1,111 @@
+"""
+Week 11 - Step 1: Face Mesh Tracking with Connections (Template 填空版)
+===================================================================
+本步驟目標：
+  1. 使用 OpenCV 開啟視訊鏡頭
+  2. 載入 MediaPipe 臉部特徵偵測器 (face_landmarker.task)
+  3. 偵測人臉，並在臉上畫出綠色的 3D 特效網格線與黃色特徵點
+"""
+import cv2
+import mediapipe as mp
+import numpy as np
+import sys
+import os
+
+# 1. 初始化攝影機
+cap = cv2.VideoCapture(0)
+if not cap.isOpened():
+    print("Error: Could not open camera.")
+    sys.exit()
+
+# 2. 載入 MediaPipe 臉部偵測器模型 (Face Landmarker)
+model_path = os.path.join('face_landmarker.task')
+if not os.path.exists(model_path):
+    print(f"Error: Model file not found at {model_path}")
+    print("Please make sure face_landmarker.task is in the examples folder.")
+    sys.exit()
+
+try:
+    from mediapipe.tasks import python
+    from mediapipe.tasks.python import vision
+    base_options = python.BaseOptions(model_asset_path=model_path)
+    options = vision.FaceLandmarkerOptions(base_options=base_options, num_faces=1)
+    detector = vision.FaceLandmarker.create_from_options(options)
+    print("Face Landmarker Tasks API initialized successfully.")
+except Exception as e:
+    print(f"Error initializing detector: {e}")
+    sys.exit()
+
+# 匯入臉部網格連接點定義
+try:
+    sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..'))
+    from face_mesh_connections import FACEMESH_TESSELATION
+except ImportError:
+    FACEMESH_TESSELATION = []
+
+# 繪圖模式：0 = 僅畫網格線, 1 = 僅畫點點, 2 = 點線都畫
+draw_mode = 0
+
+print("Press 'q' in the window to exit, or 'm' to toggle display modes.")
+
+# 3. 主迴圈
+while True:
+    success, img = cap.read()
+    if not success:
+        print("Error: Failed to read frame.")
+        break
+
+    # 【TODO 1】畫面水平翻轉，並將 BGR 轉成 MediaPipe 需要的 RGB 格式
+    img = cv2.flip(img, 1)
+    img_h, img_w, _ = img.shape
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+    
+
+    # 執行臉部偵測
+    results = detector.detect(mp_image)
+
+    # 【TODO 2】如果偵測到臉部，畫出 3D 特效網格線
+    # 提示：取出 results.face_landmarks[0]，並用迴圈巡覽 FACEMESH_TESSELATION
+    # 點的座標 (lm.x, lm.y) 是 0~1 的比例，需要乘上 img_w 和 img_h 換算成像素座標
+    if results.face_landmarks:
+        face_lms = results.face_landmarks[0]
+        num_lms = len(face_lms)
+    
+        if draw_mode == 0 or draw_mode == 2:
+            for start_idx, end_idx in FACEMESH_TESSELATION:
+                if start_idx < num_lms and end_idx < num_lms:
+                    lm1 = face_lms[start_idx]
+                    lm2 = face_lms[end_idx]
+                    pt1 = (int(lm1.x * img_w), int(lm1.y * img_h))
+                    pt2 = (int(lm2.x * img_w), int(lm2.y * img_h))
+                    cv2.line(img, pt1, pt2, (224, 224, 0), 1) # 青綠色線條
+
+
+        # 【TODO 3】畫出臉部 468 個黃色特徵點點
+        if draw_mode == 1 or draw_mode == 2:
+            for lm in face_lms:
+                cx = int(lm.x * img_w)
+                cy = int(lm.y * img_h)
+                cv2.circle(img, (cx, cy), 1, (0, 255, 255), -1) # 黃色小圓點
+
+
+
+        mode_str = "Mesh" if draw_mode == 0 else "Dots" if draw_mode == 1 else "Both"
+        cv2.putText(img, f"Face Detected | Mode: {mode_str}", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 255, 0), 2)
+    else:
+        cv2.putText(img, "No Face Detected", (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
+
+    # 顯示操作提示文字
+    cv2.putText(img, "Press 'm' to toggle mode | 'q' to quit", (10, img_h - 20), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 1)
+
+    cv2.imshow("Face Mesh Connections", img)
+
+    key = cv2.waitKey(1) & 0xFF
+    if key == ord('q'):
+        break
+    elif key == ord('m'):
+        draw_mode = (draw_mode + 1) % 3
+
+cap.release()
+cv2.destroyAllWindows()

projects/2026_AI營1/劉宗穎/脸部侦测/step2_facemesh_complete.py

@@ -0,0 +1,362 @@
+"""
+Week 11 - Face Mesh Expression Detection (Complete Template 填空版)
+===========================================================
+本步驟目標：
+  1. 將 MediaPipe Face Landmarker 與 Pygame 完美結合，打造科幻儀表板
+  2. 計算臉部指標：嘴巴張開比例與眼睛胖瘦比例 (EAR)
+  3. 實作「雙門檻滯後防吵鎖」，避免警報器反覆爆音
+  4. 實作「跳幀推理優化」，解決 AI 推理造成的畫面卡頓
+  5. 整合音效播放功能
+"""
+import pygame
+import cv2
+import mediapipe as mp
+import math
+import sys
+import os
+import time
+
+# ============================================
+# 1. 初始化 Pygame & 音效
+# ============================================
+pygame.init()
+pygame.mixer.init()
+WIDTH, HEIGHT = 800, 600
+screen = pygame.display.set_mode((WIDTH, HEIGHT))
+pygame.display.set_caption("Week 11 - Face Mesh Expression Detector ")
+clock = pygame.time.Clock()
+
+font = pygame.font.SysFont(None, 36)
+small_font = pygame.font.SysFont(None, 24)
+title_font = pygame.font.SysFont(None, 48)
+
+# 載入警告音效 (shoot.wav)
+try:
+    trigger_sound = pygame.mixer.Sound(os.path.join(os.path.dirname(__file__), "shoot.wav"))
+    print("Success: Loaded sound file shoot.wav")
+except FileNotFoundError:
+    trigger_sound = None
+    print("Warning: Could not find sound file shoot.wav")
+
+# ============================================
+# 2. 初始化 MediaPipe Face Landmarker
+# ============================================
+model_path = os.path.join(os.path.dirname(__file__), 'face_landmarker.task')
+if not os.path.exists(model_path):
+    # 若在當前資料夾找不到，就去上一層找
+    model_path = os.path.join(os.path.dirname(__file__), '..', 'face_landmarker.task')
+    if not os.path.exists(model_path):
+        print(f"Error: Model file not found at {model_path}")
+        print("Please make sure face_landmarker.task is in the directory.")
+        pygame.quit()
+        sys.exit()
+
+# 【TODO 1】初始化 MediaPipe Face Landmarker 偵測器
+# try:
+#     from mediapipe.tasks import python
+#     from mediapipe.tasks.python import vision
+#     base_options = python.BaseOptions(model_asset_path=model_path)
+#     options = vision.FaceLandmarkerOptions(base_options=base_options, num_faces=1)
+#     detector = vision.FaceLandmarker.create_from_options(options)
+#     print("Success: Loaded MediaPipe detector model")
+# except Exception as e:
+#     print(f"Error: Failed to load model: {e}")
+#     pygame.quit()
+#     sys.exit()
+
+try:
+    from mediapipe.tasks import python
+    from mediapipe.tasks.python import vision
+    base_options = python.BaseOptions(model_asset_path=model_path)
+    options = vision.FaceLandmarkerOptions(base_options=base_options, num_faces=1)
+    detector = vision.FaceLandmarker.create_from_options(options)
+    print("Success: Loaded MediaPipe detector model")
+except Exception as e:
+    print(f"Error: Failed to load model: {e}")
+    pygame.quit()
+    sys.exit()
+
+# ============================================
+# 3. 臉部特徵點索引定義
+# ============================================
+LEFT_EYE_OUTER = 33
+RIGHT_EYE_OUTER = 263
+LIP_TOP = 13
+LIP_BOTTOM = 14
+LEFT_EYE_TOP = 159
+LEFT_EYE_BOTTOM = 145
+LEFT_EYE_RIGHT = 133
+
+# ============================================
+# 4. 門檻值與狀態變數
+# ============================================
+# 嘴巴雙門檻防吵鎖
+MOUTH_OPEN_THRESHOLD = 0.35
+MOUTH_CLOSE_THRESHOLD = 0.15
+
+# 眼睛雙門檻防吵鎖
+EAR_BLINK_THRESHOLD = 0.18
+EAR_OPEN_THRESHOLD = 0.25
+
+is_mouth_open = False
+was_mouth_open = False
+
+is_blinking = False
+was_blinking = False
+
+mouth_open_count = 0
+blink_count = 0
+
+# 跳幀優化變數：每 3 影格才做一次 AI 推理
+frame_counter = 0
+FRAME_SKIP = 3
+detector_results = None
+
+# 開啟攝影機
+cap = cv2.VideoCapture(0)
+if not cap.isOpened():
+    print("Error: Could not open camera.")
+    pygame.quit()
+    sys.exit()
+
+# ============================================
+# 5. 主迴圈
+# ============================================
+running = True
+while running:
+    for event in pygame.event.get():
+        if event.type == pygame.QUIT:
+            running = False
+        elif event.type == pygame.KEYDOWN:
+            if event.key == pygame.K_ESCAPE:
+                running = False
+
+    success, img = cap.read()
+    if not success:
+        cap.set(cv2.CAP_PROP_POS_FRAMES, 0)
+        continue
+
+    img = cv2.flip(img, 1)
+    img_h, img_w, _ = img.shape
+    frame_counter += 1
+
+    plot_points = {}
+
+    # 【TODO 2】實作跳幀推理優化 (每 FRAME_SKIP 影格才跑一次 AI)
+    # if frame_counter % FRAME_SKIP == 0 or detector_results is None:
+    #     try:
+    #         img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    #         mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+    #         detector_results = detector.detect(mp_image)
+    #     except Exception as e:
+    #         print(f"Error: Inference exception: {e}")
+
+    if frame_counter % FRAME_SKIP == 0 or detector_results is None:
+        try:
+            img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+            mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+            detector_results = detector.detect(mp_image)
+        except Exception as e:
+            print(f"Error: Inference exception: {e}")
+
+    # 【TODO 3】特徵點座標映射與比例計算
+    # 1. 取得對應點在 Pygame 視窗 (WIDTH, HEIGHT) 中的像素位置並存入 plot_points
+    # 2. 計算眼寬基準線 base_dist
+    # 3. 計算歸一化嘴巴比例 mouth_ratio
+    # 4. 計算眼睛胖瘦比 ear (提示：垂直高度 eye_h / 水平寬度 eye_w)
+    # if detector_results and detector_results.face_landmarks:
+    #     face_lms = detector_results.face_landmarks[0]
+    #
+    #     for idx in [LEFT_EYE_OUTER, RIGHT_EYE_OUTER, LIP_TOP, LIP_BOTTOM, LEFT_EYE_TOP, LEFT_EYE_BOTTOM, LEFT_EYE_RIGHT]:
+    #         pt = face_lms[idx]
+    #         plot_points[idx] = (int(pt.x * WIDTH), int(pt.y * HEIGHT))
+    #
+    #     p_lx, p_ly = plot_points[LEFT_EYE_OUTER]
+    #     p_rx, p_ry = plot_points[RIGHT_EYE_OUTER]
+    #     base_dist = math.hypot(p_rx - p_lx, p_ry - p_ly)
+    #
+    #     if base_dist > 0:
+    #         p_m_top_x, p_m_top_y = plot_points[LIP_TOP]
+    #         p_m_bot_x, p_m_bot_y = plot_points[LIP_BOTTOM]
+    #         mouth_dist = math.hypot(p_m_bot_x - p_m_top_x, p_m_bot_y - p_m_top_y)
+    #         mouth_ratio = mouth_dist / base_dist
+    #
+    #         p_e_top_x, p_e_top_y = plot_points[LEFT_EYE_TOP]
+    #         p_e_bot_x, p_e_bot_y = plot_points[LEFT_EYE_BOTTOM]
+    #         p_e_right_x, p_e_right_y = plot_points[LEFT_EYE_RIGHT]
+    #
+    #         eye_h = math.hypot(p_e_bot_x - p_e_top_x, p_e_bot_y - p_e_top_y)
+    #         eye_w = math.hypot(p_e_right_x - p_lx, p_e_right_y - p_ly)
+    #         ear = eye_h / eye_w if eye_w > 0 else 0
+
+    if detector_results and detector_results.face_landmarks:
+        face_lms = detector_results.face_landmarks[0]
+
+        for idx in [LEFT_EYE_OUTER, RIGHT_EYE_OUTER, LIP_TOP, LIP_BOTTOM, LEFT_EYE_TOP, LEFT_EYE_BOTTOM, LEFT_EYE_RIGHT]:
+            pt = face_lms[idx]
+            plot_points[idx] = (int(pt.x * WIDTH), int(pt.y * HEIGHT))
+
+        p_lx, p_ly = plot_points[LEFT_EYE_OUTER]
+        p_rx, p_ry = plot_points[RIGHT_EYE_OUTER]
+        base_dist = math.hypot(p_rx - p_lx, p_ry - p_ly)
+
+        if base_dist > 0:
+            p_m_top_x, p_m_top_y = plot_points[LIP_TOP]
+            p_m_bot_x, p_m_bot_y = plot_points[LIP_BOTTOM]
+            mouth_dist = math.hypot(p_m_bot_x - p_m_top_x, p_m_bot_y - p_m_top_y)
+            mouth_ratio = mouth_dist / base_dist
+
+            p_e_top_x, p_e_top_y = plot_points[LEFT_EYE_TOP]
+            p_e_bot_x, p_e_bot_y = plot_points[LEFT_EYE_BOTTOM]
+            p_e_right_x, p_e_right_y = plot_points[LEFT_EYE_RIGHT]
+
+            eye_h = math.hypot(p_e_bot_x - p_e_top_x, p_e_bot_y - p_e_top_y)
+            eye_w = math.hypot(p_e_right_x - p_lx, p_e_right_y - p_ly)
+            ear = eye_h / eye_w if eye_w > 0 else 0
+
+            # 【TODO 4】實作雙門檻防吵鎖 (滯後判斷邏輯)
+            # if mouth_ratio > MOUTH_OPEN_THRESHOLD:
+            #     is_mouth_open = True
+            # elif mouth_ratio < MOUTH_CLOSE_THRESHOLD:
+            #     is_mouth_open = False
+            #
+            # if ear < EAR_BLINK_THRESHOLD:
+            #     is_blinking = True
+            # elif ear > EAR_OPEN_THRESHOLD:
+            #     is_blinking = False
+
+            if mouth_ratio > MOUTH_OPEN_THRESHOLD:
+                is_mouth_open = True
+            elif mouth_ratio < MOUTH_CLOSE_THRESHOLD:
+                is_mouth_open = False
+
+            if ear < EAR_BLINK_THRESHOLD:
+                is_blinking = True
+            elif ear > EAR_OPEN_THRESHOLD:
+                is_blinking = False
+
+            # 【TODO 5】實作邊緣觸發判定 (只有在剛張嘴或剛閉眼的那一瞬間，播放一次警告音效)
+            # if is_mouth_open and not was_mouth_open:
+            #     mouth_open_count += 1
+            #     if trigger_sound:
+            #         trigger_sound.play()
+            #
+            # if is_blinking and not was_blinking:
+            #     blink_count += 1
+            #
+            # was_mouth_open = is_mouth_open
+            # was_blinking = is_blinking
+
+            if is_mouth_open and not was_mouth_open:
+                mouth_open_count += 1
+                if trigger_sound:
+                    trigger_sound.play()
+
+            if is_blinking and not was_blinking:
+                blink_count += 1
+
+            was_mouth_open = is_mouth_open
+            was_blinking = is_blinking
+
+    else:
+        is_mouth_open = False
+        is_blinking = False
+        was_mouth_open = False
+        was_blinking = False
+        mouth_ratio = 0.0
+        ear = 0.30
+
+    # ============================================
+    # 6. 渲染 Pygame 畫面與 UI 元素
+    # ============================================
+    # 【TODO 6】根據狀態切換背景顏色 (安全時深色，張嘴時深綠色，眨眼時深藍色)
+    # if is_mouth_open:
+    #     screen.fill((16, 44, 28))
+    # elif is_blinking:
+    #     screen.fill((15, 30, 54))
+    # else:
+    #     screen.fill((20, 20, 30))
+
+    if is_mouth_open:
+        screen.fill((16, 44, 28))
+    elif is_blinking:
+        screen.fill((15, 30, 54))
+    else:
+        screen.fill((20, 20, 30))
+
+    # 【TODO 7】畫出眼寬基準線、眼睛高度線與嘴唇高度線
+    # 提示：用 pygame.draw.line 在螢幕上將點位連線
+    # if plot_points:
+    #     pygame.draw.line(screen, (100, 100, 150), plot_points[LEFT_EYE_OUTER], plot_points[RIGHT_EYE_OUTER], 2)
+    #     pygame.draw.line(screen, (245, 158, 11), plot_points[LIP_TOP], plot_points[LIP_BOTTOM], 3)
+    #     pygame.draw.line(screen, (59, 130, 246), plot_points[LEFT_EYE_TOP], plot_points[LEFT_EYE_BOTTOM], 3)
+
+    if plot_points:
+        pygame.draw.line(screen, (100, 100, 150), plot_points[LEFT_EYE_OUTER], plot_points[RIGHT_EYE_OUTER], 2)
+        pygame.draw.line(screen, (245, 158, 11), plot_points[LIP_TOP], plot_points[LIP_BOTTOM], 3)
+        pygame.draw.line(screen, (59, 130, 246), plot_points[LEFT_EYE_TOP], plot_points[LEFT_EYE_BOTTOM], 3)
+
+        # 畫特徵點小圓圈
+        for idx, pos in plot_points.items():
+            color = (255, 255, 255)
+            if idx in [LIP_TOP, LIP_BOTTOM]:
+                color = (245, 158, 11)
+            elif idx in [LEFT_EYE_TOP, LEFT_EYE_BOTTOM, LEFT_EYE_OUTER, LEFT_EYE_RIGHT]:
+                color = (59, 130, 246)
+            pygame.draw.circle(screen, color, pos, 6)
+
+    # 繪製儀表板 HUD 卡片
+    pygame.draw.rect(screen, (30, 30, 45), (30, 30, 340, 220), border_radius=12)
+    pygame.draw.rect(screen, (50, 50, 70), (30, 30, 340, 220), width=2, border_radius=12)
+
+    screen.blit(font.render("Face Mesh Dashboard", True, (255, 255, 255)), (50, 45))
+
+    m_color = (34, 197, 94) if is_mouth_open else (245, 158, 11)
+    screen.blit(small_font.render(f"Mouth Ratio: {mouth_ratio:.2f}", True, m_color), (50, 90))
+    pygame.draw.rect(screen, (50, 50, 65), (50, 110, 200, 12), border_radius=4)
+    pygame.draw.rect(screen, m_color, (50, 110, int(min(mouth_ratio, 1.0) * 200), 12), border_radius=4)
+
+    e_color = (59, 130, 246) if not is_blinking else (239, 68, 68)
+    screen.blit(small_font.render(f"Eye EAR: {ear:.2f}", True, e_color), (50, 140))
+    pygame.draw.rect(screen, (50, 50, 65), (50, 160, 200, 12), border_radius=4)
+    pygame.draw.rect(screen, e_color, (50, 160, int(min(ear, 0.5) * 400), 12), border_radius=4)
+
+    # 繪製閾值指示白線
+    pygame.draw.line(screen, (255, 255, 255), (50 + int(MOUTH_OPEN_THRESHOLD * 200), 108), (50 + int(MOUTH_OPEN_THRESHOLD * 200), 124), 2)
+    pygame.draw.line(screen, (255, 255, 255), (50 + int(EAR_BLINK_THRESHOLD * 400), 158), (50 + int(EAR_BLINK_THRESHOLD * 400), 174), 2)
+
+    # 繪製統計卡片
+    pygame.draw.rect(screen, (30, 30, 45), (WIDTH - 320, 30, 290, 110), border_radius=12)
+    pygame.draw.rect(screen, (50, 50, 70), (WIDTH - 320, 30, 290, 110), width=2, border_radius=12)
+    screen.blit(small_font.render(f"Mouth Open Count: {mouth_open_count}", True, (255, 255, 255)), (WIDTH - 300, 45))
+    screen.blit(small_font.render(f"Blink/Close Count: {blink_count}", True, (255, 255, 255)), (WIDTH - 300, 75))
+    screen.blit(small_font.render(f"Hysteresis Status: ACTIVE", True, (34, 197, 94)), (WIDTH - 300, 105))
+
+    # 顯示置中警告狀態大字
+    if is_mouth_open:
+        status_lbl = title_font.render("MOUTH OPENED", True, (34, 197, 94))
+        screen.blit(status_lbl, (WIDTH // 2 - status_lbl.get_width() // 2, HEIGHT // 2 + 100))
+    elif is_blinking:
+        status_lbl = title_font.render("BLINKING", True, (59, 130, 246))
+        screen.blit(status_lbl, (WIDTH // 2 - status_lbl.get_width() // 2, HEIGHT // 2 + 100))
+    else:
+        if not plot_points:
+            status_lbl = title_font.render("NO FACE DETECTED", True, (239, 68, 68))
+            screen.blit(status_lbl, (WIDTH // 2 - status_lbl.get_width() // 2, HEIGHT // 2 - 20))
+        else:
+            status_lbl = title_font.render("NORMAL", True, (200, 200, 200))
+            screen.blit(status_lbl, (WIDTH // 2 - status_lbl.get_width() // 2, HEIGHT // 2 + 100))
+
+    # 顯示底部效能狀態文字
+    status_text = small_font.render(
+        f"Frame: {frame_counter} | Skip: {FRAME_SKIP} frames | FPS: {clock.get_fps():.0f}",
+        True, (100, 100, 120)
+    )
+    screen.blit(status_text, (20, HEIGHT - 35))
+
+    pygame.display.update()
+    clock.tick(30)
+
+cap.release()
+pygame.quit()
+sys.exit()

projects/2026_AI營1/劉宗穎/设计AI角色&场景/d2_2_prompt_engineering.py

@@ -0,0 +1,56 @@
+# d2_2_prompt_engineering.py
+# Day 2: 生成式 AI 串接實戰 - Prompt 工程 (系統指令與結構化輸出)
+import os
+from google import genai
+
+# 取得 API 金鑰
+api_key = os.environ.get("GEMINI_API_KEY", "AQ.Ab8RN6JvbUqLW-wcddeMducmta0aGE2QtcrmF4Kfd8nnpZ9mzg")
+if not api_key:
+    print("[提示] 請輸入您的 Google AI Studio API 金鑰：")
+    api_key = input("API Key: ").strip()
+
+client = genai.Client(api_key=api_key)
+
+# 1. 設置系統指令 (System Instruction / System Prompt)
+# TODO 1:
+system_prompt = "你是AI國中英語老師，请输入你想要翻译的中文单字"
+
+print("正在初始化强制英語教學模型...")
+
+# 2. Few-shot (少樣本) 引導與對話演練
+# TODO 2:
+print(f"\n[學生輸入]: ")
+user_word = input().strip()
+prompt = f"請幫我翻譯這個詞：'{user_word}'。"
+
+response = client.models.generate_content(
+    model='gemini-3.1-flash-lite',
+    contents=prompt,
+    config={'system_instruction': system_prompt}
+)
+
+print("\n[AI 老師回應]:")
+print(response.text)
+
+
+# 3. 結構化 JSON 輸出 (Structured Output) 範例
+print("\n" + "="*40)
+print("進階挑戰：要求模型回傳 JSON 格式")
+print("="*40)
+
+# 我們可以在 prompt 中要求，並透過 configuration 設定 JSON 輸出格式
+# TODO 3:
+json_prompt = (
+    f"請將中文單字 '{user_word}' 的翻譯結果整理成 JSON 格式，包含以下三個欄位：\n"
+    "- 'english': 英文單字\n"
+    "- 'kk_phonetic': KK音標\n"
+    "- 'example_sentence': 適合國中生理解的英文例句與中文翻譯"
+)
+
+response = client.models.generate_content(
+    model='gemini-3.1-flash-lite',
+    contents=json_prompt,
+    config={"response_mime_type": "application/json"}
+)
+print("\n[AI 回傳的原始 JSON]:")
+print(response.text)

projects/2026_AI營1/劉宗穎/资料异常值清洗/d1_2_data_cleaning.py

@@ -0,0 +1,38 @@
+# d1_2_data_cleaning.py
+# Day 1 學員填空：資料清洗（完成 TODO）
+# cd examples/Day1/student && python3 d1_2_data_cleaning.py
+# 參考答案：../Teacher/d1_2_data_cleaning.py
+
+import pandas as pd
+import numpy as np
+
+# 與 Teacher 一致：第 5 天高溫有雨，方便篩選示範
+raw_data = {
+    "Date": ["2026-07-01", "2026-07-02", "2026-07-03", "2026-07-04", "2026-07-05"],
+    "Temperature": [28.5, np.nan, 30.2, 31.5, 30.8],
+    "Humidity": [82, 80, np.nan, 70, 90],
+    "Rainfall": [5.2, 0.0, 0.0, np.nan, 25.4],
+}
+
+df = pd.DataFrame(raw_data)
+print("=== 原始含有缺失值的數據 ===")
+print(df)
+
+df_cleaned = df.copy()
+mean_temp = df_cleaned["Temperature"].mean()
+df_cleaned["Temperature"] = df_cleaned["Temperature"].fillna(mean_temp)
+df_cleaned["Humidity"] = df_cleaned["Humidity"].ffill()
+df_cleaned["Rainfall"] = df_cleaned["Rainfall"].fillna(0.0)
+
+print("\n=== 清洗與填補後的數據 ===")
+print(df_cleaned)
+
+print("\n=== 篩選：溫度>29 且 降雨>0 ===")
+high_temp_and_rain = df_cleaned[(df_cleaned['Temperature'] > 29) & (df_cleaned['Rainfall'] > 0)]
+print(high_temp_and_rain)
+
+print("\n=== 新增 Is_Hot（溫度 >= 30）===")
+df_cleaned['Is_Hot'] = df_cleaned["Temperature"] >= 30
+print(df_cleaned)
+df_cleaned['rainy'] = df_cleaned["Rainfall"] > 0
+print(df_cleaned)

projects/2026_AI營1/劉宗穎/资料异常值清洗/taipei_weather_2026.csv

@@ -0,0 +1,16 @@
+Date,Temperature,Humidity,Rainfall
+2026-07-01,28.5,82,5.2
+2026-07-02,29.0,80,0.0
+2026-07-03,30.2,75,0.0
+2026-07-04,31.5,70,0.0
+2026-07-05,27.8,90,25.4
+2026-07-06,26.5,95,42.0
+2026-07-07,28.0,85,12.5
+2026-07-08,29.5,78,0.0
+2026-07-09,30.8,72,0.0
+2026-07-10,32.0,68,0.0
+2026-07-11,28.2,88,18.0
+2026-07-12,27.0,92,30.5
+2026-07-13,29.2,80,2.1
+2026-07-14,30.0,76,0.0
+2026-07-15,31.2,71,0.0

projects/2026_AI營1/劉柔佑/DataFrame指令練習/d1_1_pandas_intro.py

@@ -0,0 +1,34 @@
+# d1_1_pandas_intro.py
+# Day 1 學員填空：Pandas 基礎（完成 TODO 後執行）
+# cd examples/Day1/student && python3 d1_1_pandas_intro.py
+# 參考答案：../Teacher/d1_1_pandas_intro.py
+
+import pandas as pd
+
+df = pd.read_csv("taipei_weather_2026.csv")
+print("=== 數據載入成功！ ===")
+
+print("\n--- 1. head ---")
+print(df.head())
+
+print("\n--- 2. info ---")
+df.info()
+
+print("\n--- 3. describe ---")
+print(df.describe())
+
+print("\n--- 4. Temperature 前 5 筆 ---")
+print(df['Temperature'].head())
+
+print("\n--- 5. Date 與 Rainfall ---")
+print(df[['Date', 'Rainfall']].head())
+
+print("\n--- 挑戰：最高溫與平均濕度 ---")
+max_temp = df['Temperature'].max()
+avg_humidity = df['Humidity'].mean()
+print(f"最高溫度：{max_temp}°C")
+print(f"平均濕度：{avg_humidity:.2f}%")
+min_Rainfall = df['Rainfall'].min()
+print(f"平均濕度：{min_Rainfall:.2f}%")
+egyptian_good_weather =df[(df["Temperature"]>27) & (df["Rainfall"]==0)]
+print(egyptian_good_weather)

projects/2026_AI營1/劉柔佑/DataFrame指令練習/taipei_weather_2026.csv

@@ -0,0 +1,16 @@
+Date,Temperature,Humidity,Rainfall
+2026-07-01,28.5,82,5.2
+2026-07-02,29.0,80,0.0
+2026-07-03,30.2,75,0.0
+2026-07-04,31.5,70,0.0
+2026-07-05,27.8,90,25.4
+2026-07-06,26.5,95,42.0
+2026-07-07,28.0,85,12.5
+2026-07-08,29.5,78,0.0
+2026-07-09,30.8,72,0.0
+2026-07-10,32.0,68,0.0
+2026-07-11,28.2,88,18.0
+2026-07-12,27.0,92,30.5
+2026-07-13,29.2,80,2.1
+2026-07-14,30.0,76,0.0
+2026-07-15,31.2,71,0.0

projects/2026_AI營1/劉柔佑/ai_api串接練習/d2_1_gemini.py

@@ -0,0 +1,41 @@
+# d2_1_gemini.py — Day2 學員填空
+# cd examples/Day2/Student && python3 d2_1_gemini.py
+# 參考答案：../Teacher/d2_1_gemini.py
+
+import os
+from google import genai
+
+# === 【學員填空練習點 1：設置與載入 API Key】 ===
+# TODO: 嘗試從環境變數 GEMINI_API_KEY 取得 API Key。如果未設定，則使用 input() 提示手動輸入。
+api_key = os.environ.get("GEMINI_API_KEY", "AQ.Ab8RN6J-S5dtl4YCuomoO9V4L5YAgTzLh9E8vJUuScN0c4VyKg")
+
+if not api_key:
+    print("[提示] 未檢測到環境變數 GEMINI_API_KEY。")
+    print("      請先設定：export GEMINI_API_KEY=\"your_key\"") 
+    print("      或於下方手動輸入（僅供本機測試，勿寫死在檔案中上傳）：")
+    api_key = input("API Key: ").strip()
+    if not api_key:
+        print("[結束] 未提供 API Key。可改練習 d2_1_ollama_local.py。")
+        exit(0)
+
+# === 【學員填空練習點 2：設定模型名稱】 ===
+# TODO: 設定使用的 Gemini 模型名稱（例如 'gemini-3.1-flash-lite'）
+model_name = "gemini-3.1-flash-lite"
+
+print(f"正在連線至 Google Cloud Model ({model_name})...")
+
+# === 【學員填空練習點 3：初始化 Gemini Client】 ===
+# TODO: 呼叫 genai.Client 並傳入 api_key 參數來初始化 Client 物件，指派給 client 變數
+client = genai.Client(api_key=api_key)
+
+# === 【學員填空練習點 4：發送內容生成請求】 ===
+# TODO: 使用 client.models.generate_content 呼叫模型，傳入 model 參數與 contents 提示詞
+#       提示詞內容：請模型以繁體中文解釋什麼是『API 串接』，並用餐廳點餐作為比喻，字數約 100 字。
+response = client.models.generate_content(model= model_name, contents="請以繁體中文解釋什麼是『AI』，字數約 100 字。")
+
+
+# === 【學員填空練習點 5：讀取並印出回應】 ===
+# TODO: 從 response 物件中取得 text 欄位內容，並列印出來。
+print("\n=== AI 回應結果 ===")
+# 請在下方實作列印邏輯：
+print(response.text)

projects/2026_AI營1/劉柔佑/draw_game/draw_game_student.py

@@ -0,0 +1,147 @@
+import cv2
+import numpy as np
+import mediapipe as mp
+from mediapipe.tasks import python
+from mediapipe.tasks.python import vision
+
+# ============================================
+# Week 9 - 簡單學生版：支援多色切換的 AR 空中寫字板 (純黑畫布版)
+# ============================================
+
+# === 【重要觀念 TODO 1：MediaPipe 手部偵測模型初始化】 ===
+# 觀念重點：
+# 1. BaseOptions 用於設定模型檔案的路徑，這裡使用下載好的 'hand_landmarker.task'。
+# 2. HandLandmarkerOptions 用於設定手部偵測的細項設定，例如 num_hands=1 代表最多偵測一隻手。
+# 3. 最後透過 vision.HandLandmarker.create_from_options 建立手部偵測器物件。
+# 請解開下方三行程式碼的註釋：
+base_options = python.BaseOptions(model_asset_path='hand_landmarker.task')
+options = vision.HandLandmarkerOptions(base_options=base_options, num_hands=1)
+detector = vision.HandLandmarker.create_from_options(options)
+
+cap = cv2.VideoCapture(0)
+canvas = None
+px, py = 0, 0
+
+# 定義幾種高對比的顏色 (格式為 B, G, R)
+COLORS = {
+    'purple': (191,161,196),     # 紫色
+    'turquoise': (128,217,255),    # 青綠色
+    'skyblue': (138,202,229),  # 天蓝色
+    'pink': (213,110,237)        # 粉色
+}
+
+# 預設使用螢光綠
+current_color = COLORS['purple']
+
+print("系統啟動中...")
+print("【操作指南】")
+print("☝️  單伸食指：下筆寫字")
+print("✌️  食指+中指：懸停移動 (不寫字)")
+print("⌨️  按下 '1~4'：切換畫筆顏色")
+print("⌨️  按下 's'：清空畫面")
+print("⌨️  按下 'f'：結束程式")
+
+while True:
+    success, img = cap.read()
+    if not success:
+        break
+
+    # === 【重要觀念 TODO 2：攝影機影像水平翻轉】 ===
+    # 觀念重點：
+    # 1. 攝影機預設取得的影像與我們的真實動作是相反的（非鏡像）。
+    # 2. 在互動式 AR 應用中，為了讓使用者的左手移動時，畫面中的手也往左移動，
+    #    必須使用 cv2.flip(img, 1) 將影像水平翻轉，以提供符合直覺的「鏡像體驗」。
+    # 請解開下方一行程式碼的註釋：
+    img = cv2.flip(img, 1)
+
+    h, w, c = img.shape
+
+    # 初始化畫布
+    if canvas is None:
+        canvas = np.zeros((h, w, c), dtype=np.uint8)
+
+    # 為了維持畫筆提示點（或懸停圓圈）的動態顯示，且不破壞原本的 canvas 紀錄，
+    # 我們在每一訊框都建立一個純黑的暫時畫布來呈現最終畫面。
+    display_img = np.zeros((h, w, c), dtype=np.uint8)
+
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=img_rgb)
+    results = detector.detect(mp_image)
+
+    if results.hand_landmarks:
+        hand_lms = results.hand_landmarks[0]
+
+        index_tip = hand_lms[8]
+        index_pip = hand_lms[6]
+        middle_tip = hand_lms[12]
+        middle_pip = hand_lms[10]
+
+        cx, cy = int(index_tip.x * w), int(index_tip.y * h)
+
+        index_up = index_tip.y < index_pip.y
+        middle_up = middle_tip.y < middle_pip.y
+
+        if index_up and not middle_up:
+            # 狀態：【繪圖模式】
+            if px == 0 and py == 0:
+                px, py = cx, cy
+
+            # === 【重要觀念 TODO 3：在 OpenCV Canvas 上繪製軌跡】 ===
+            # 觀念重點：
+            # 1. cv2.line 用於在影像上繪製直線。
+            # 2. 參數說明：畫布(canvas)、起點(px, py)、終點(cx, cy)、顏色(current_color)、線寬(8)。
+            # 3. 藉由在每一訊框連續繪製前一個位置到當前位置的線段，即可形成平滑的寫字軌跡。
+            # 請解開下方一行程式碼的註釋：
+            cv2.line(canvas, (px, py), (cx, cy), current_color, 8)
+            
+            px, py = cx, cy
+
+            # 筆頭提示點直接畫在暫時的顯示畫布上，避免在黑畫布上找不到筆尖
+            cv2.circle(display_img, (cx, cy), 10, current_color, cv2.FILLED)
+
+        elif index_up and middle_up:
+            # 狀態：【懸停模式】
+            px, py = 0, 0
+            # 懸停提示圈畫在暫時的顯示畫布上，移動時才不會在黑背景中迷路
+            cv2.circle(display_img, (cx, cy), 12, (255, 255, 255), 2)
+
+        else:
+            px, py = 0, 0
+    else:
+        px, py = 0, 0
+
+    # 為什麼改用 canvas 與 display_img 疊加？
+    # 原本是與相機影像 (img) 疊加，現在改與純黑的 display_img 疊加，
+    # 這樣畫面上就只會留下寫字的軌跡以及當前的筆頭/懸停提示，完全遮蔽人像背景。
+    
+    # === 【重要觀念 TODO 4：AR 畫面疊加與顯示】 ===
+    # 觀念重點：
+    # 1. cv2.add 用於將兩個相同大小與類型的影像矩陣進行相加。
+    # 2. 這裡將寫字痕跡畫布 (canvas) 與當前的筆頭/懸停提示畫布 (display_img) 疊加。
+    # 3. 疊加後使用 cv2.imshow 在視窗中顯示最終成果。
+    # 請解開下方兩行程式碼的註釋：
+    final_output = cv2.add(canvas, display_img)
+    cv2.imshow("Hand Tracking - AR Canvas", final_output)
+
+    # 鍵盤監聽與控制邏輯
+    key = cv2.waitKey(1) & 0xFF
+    if key == ord('q'):
+        break
+    elif key == ord('c'):
+        canvas = np.zeros((h, w, c), dtype=np.uint8)
+
+    elif key == ord('1'):
+        current_color = COLORS['purple']
+        print("已切換為：紫色")
+    elif key == ord('2'):
+        current_color = COLORS['turquoise']
+        print("已切換為：青綠色")
+    elif key == ord('3'):
+        current_color = COLORS['skyblue']
+        print("已切換為：蓝色")
+    elif key == ord('4'):
+        current_color = COLORS['pink']
+        print("已切換為：粉色")
+
+cap.release()
+cv2.destroyAllWindows()