| import gradio as gr |
| from transformers import pipeline |
| import spaces |
|
|
| |
| classifier = None |
|
|
| @spaces.GPU |
| def predict_plant(image): |
| global classifier |
| if image is None: |
| return None |
|
|
| try: |
| if classifier is None: |
| print("正在加载模型 (Swin Tiny，绝对可用版)...") |
| |
| classifier = pipeline( |
| task="image-classification", |
| model="microsoft/swin-tiny-patch4-window7-224", |
| top_k=5 |
| ) |
| print("模型加载完成！") |
|
|
| results = classifier(image) |
| |
| formatted_results = {} |
| for res in results: |
| if res["score"] > 0.001: |
| formatted_results[res["label"]] = float(res["score"]) |
| |
| if not formatted_results: |
| return {"未能识别具体植物": 1.0} |
| |
| return formatted_results |
|
|
| except Exception as e: |
| return f"发生错误: {str(e)}" |
|
|
|
|
| demo = gr.Interface( |
| fn=predict_plant, |
| |
| inputs=gr.Image(type="pil", sources=["upload", "webcam"], label="🌿 请拍植物叶片特写 (避开泥土和花盆)"), |
| outputs=gr.Label(num_top_classes=5, label="🔍 视觉识别结果 (Top 5)"), |
| title="🌱 植物视觉识别 (微软 Swin Tiny)" |
| ) |
|
|
| if __name__ == "__main__": |
| demo.launch() |
