Robsonfll commited on
Commit
a80a0e4
·
verified ·
1 Parent(s): 8e9955d

Update src/streamlit_app.py

Browse files
Files changed (1) hide show
  1. src/streamlit_app.py +36 -23
src/streamlit_app.py CHANGED
@@ -2,72 +2,85 @@ import os
2
  import streamlit as st
3
  import numpy as np
4
  from PIL import Image
5
- import tensorflow as tf
6
  import json
 
7
 
8
  # ================================
9
- # Caminhos dos arquivos
10
  # ================================
11
  BASE_DIR = os.path.dirname(__file__)
12
- LABELS_PATH = os.path.join(BASE_DIR, "labels.txt")
13
- META_PATH = os.path.join(BASE_DIR, "meta.json")
14
 
15
- # 🔵 Trocaremos pelo nome do SEU modelo
16
- MODEL_PATH = os.path.join(BASE_DIR, "model.h5")
 
17
 
18
  # ================================
19
- # Carregar labels
20
  # ================================
21
  with open(LABELS_PATH, "r", encoding="utf-8") as f:
22
  class_names = [line.strip() for line in f.readlines()]
23
 
24
  # ================================
25
- # Carregar metadados
26
  # ================================
27
  with open(META_PATH, "r", encoding="utf-8") as f:
28
  meta = json.load(f)
29
 
30
- IMG_SIZE = meta["input_shape"]["height"]
31
 
32
  # ================================
33
- # Carregar modelo TensorFlow normal
34
  # ================================
35
- model = tf.keras.models.load_model(MODEL_PATH)
 
 
 
 
36
 
37
  # ================================
38
- # Função de inferência
39
  # ================================
40
  def predict(image: Image.Image):
41
  img = image.resize((IMG_SIZE, IMG_SIZE))
42
  img = np.array(img).astype("float32") / 255.0
43
  img = np.expand_dims(img, axis=0)
44
-
45
- preds = model.predict(img)[0]
46
- return preds
 
 
 
47
 
48
  # ================================
49
- # Interface Streamlit
50
  # ================================
51
  st.markdown("<h1 style='text-align:center; color:#1E88E5;'>🚀 Equipe Netuno</h1>", unsafe_allow_html=True)
52
  st.markdown("<h2 style='text-align:center;'>🗿 ArqueoIA</h2>", unsafe_allow_html=True)
53
- st.markdown("<p style='text-align:center;'>Sistema educacional de IA para identificar e classificar artefatos arqueológicos (osso, cerâmica, moeda, flecha).</p>", unsafe_allow_html=True)
54
  st.write("---")
55
 
56
- uploaded_file = st.file_uploader("Envie uma imagem do artefato:", type=["jpg", "jpeg", "png"])
 
 
 
57
 
 
 
 
58
  if uploaded_file:
59
  image = Image.open(uploaded_file).convert("RGB")
60
  st.image(image, caption="Imagem enviada", use_column_width=True)
61
 
62
  st.write("### 🔍 Analisando imagem...")
 
63
  preds = predict(image)
64
 
65
  top_idx = int(np.argmax(preds))
 
66
  top_label = class_names[top_idx]
67
- top_conf = preds[top_idx] * 100
68
 
69
- st.success(f"**Sugestão da IA: {top_label} ({top_conf:.1f}% de confiança)**")
70
 
71
- st.write("### 📊 Detalhamento das classes")
72
- for label, prob in zip(class_names, preds):
73
- st.write(f"- **{label}:** {prob*100:.1f}%")
 
2
  import streamlit as st
3
  import numpy as np
4
  from PIL import Image
 
5
  import json
6
+ import tflite_runtime.interpreter as tflite
7
 
8
  # ================================
9
+ # 1) Caminhos dos arquivos
10
  # ================================
11
  BASE_DIR = os.path.dirname(__file__)
 
 
12
 
13
+ LABELS_PATH = os.path.join(BASE_DIR, "labels.txt")
14
+ META_PATH = os.path.join(BASE_DIR, "meta.json")
15
+ MODEL_PATH = os.path.join(BASE_DIR, "model.tflite")
16
 
17
  # ================================
18
+ # 2) Carregar labels
19
  # ================================
20
  with open(LABELS_PATH, "r", encoding="utf-8") as f:
21
  class_names = [line.strip() for line in f.readlines()]
22
 
23
  # ================================
24
+ # 3) Carregar meta.json
25
  # ================================
26
  with open(META_PATH, "r", encoding="utf-8") as f:
27
  meta = json.load(f)
28
 
29
+ IMG_SIZE = meta["input_shape"]["height"] # usa altura como tamanho quadrado
30
 
31
  # ================================
32
+ # 4) Carregar modelo TFLite
33
  # ================================
34
+ interpreter = tflite.Interpreter(model_path=MODEL_PATH)
35
+ interpreter.allocate_tensors()
36
+
37
+ input_details = interpreter.get_input_details()
38
+ output_details = interpreter.get_output_details()
39
 
40
  # ================================
41
+ # 5) Função de inferência
42
  # ================================
43
  def predict(image: Image.Image):
44
  img = image.resize((IMG_SIZE, IMG_SIZE))
45
  img = np.array(img).astype("float32") / 255.0
46
  img = np.expand_dims(img, axis=0)
47
+
48
+ interpreter.set_tensor(input_details[0]["index"], img)
49
+ interpreter.invoke()
50
+
51
+ output = interpreter.get_tensor(output_details[0]["index"])[0]
52
+ return output
53
 
54
  # ================================
55
+ # 6) Interface
56
  # ================================
57
  st.markdown("<h1 style='text-align:center; color:#1E88E5;'>🚀 Equipe Netuno</h1>", unsafe_allow_html=True)
58
  st.markdown("<h2 style='text-align:center;'>🗿 ArqueoIA</h2>", unsafe_allow_html=True)
59
+ st.markdown("<p style='text-align:center;'>Sistema educacional de IA para classificar artefatos arqueológicos.</p>", unsafe_allow_html=True)
60
  st.write("---")
61
 
62
+ uploaded_file = st.file_uploader(
63
+ "Envie uma imagem do artefato:",
64
+ type=["jpg", "jpeg", "png"]
65
+ )
66
 
67
+ # ================================
68
+ # 7) Execução
69
+ # ================================
70
  if uploaded_file:
71
  image = Image.open(uploaded_file).convert("RGB")
72
  st.image(image, caption="Imagem enviada", use_column_width=True)
73
 
74
  st.write("### 🔍 Analisando imagem...")
75
+
76
  preds = predict(image)
77
 
78
  top_idx = int(np.argmax(preds))
79
+ top_conf = float(preds[top_idx]) * 100
80
  top_label = class_names[top_idx]
 
81
 
82
+ st.success(f"**Sugestão: {top_label} ({top_conf:.1f}% de confiança)**")
83
 
84
+ st.write("### 📊 Detalhamento completo")
85
+ for label, p in zip(class_names, preds):
86
+ st.write(f"- **{label}:** {p*100:.1f}%")