Add real Hugging Face model support with Food101 classification

app.py

@@ -10,6 +10,7 @@ import logging
 import time
 from datetime import datetime
 import os # Added for environment detection
+import torch # Added for Hugging Face model inference
 
 # 設置詳細日誌 - 在 Hugging Face Spaces 中只使用 StreamHandler
 logging.basicConfig(
@@ -165,14 +166,67 @@ def simulate_ai_analysis(image_url: str):
     }
 
 def real_ai_analysis(image_url: str):
-    """真實 AI 分析 - 需要模型文件"""
+    """真實 AI 分析 - 使用 Hugging Face 模型"""
     try:
-        log_analysis_step("模型載入", "嘗試載入真實 AI 模型...")
+        log_analysis_step("模型載入", "嘗試載入 Hugging Face 模型...")
         
-        # 這裡可以添加真實的模型載入代碼
-        # 由於 Docker Space 限制，這裡使用模擬模式
-        log_analysis_step("模型載入", "模型文件不可用，切換到模擬模式")
-        return simulate_ai_analysis(image_url)
+        # 嘗試載入 Hugging Face 模型
+        try:
+            from transformers import AutoImageProcessor, AutoModelForImageClassification
+            from PIL import Image
+            import requests
+            from io import BytesIO
+            
+            log_analysis_step("模型載入", "載入 Food101 模型...")
+            
+            # 載入預訓練的 Food101 模型
+            processor = AutoImageProcessor.from_pretrained("nateraw/food101")
+            model = AutoModelForImageClassification.from_pretrained("nateraw/food101")
+            
+            log_analysis_step("圖片處理", "下載並處理圖片...")
+            
+            # 下載圖片
+            response = requests.get(image_url)
+            image = Image.open(BytesIO(response.content))
+            
+            # 處理圖片
+            inputs = processor(image, return_tensors="pt")
+            
+            log_analysis_step("模型推理", "進行食物識別...")
+            
+            # 進行預測
+            with torch.no_grad():
+                outputs = model(**inputs)
+                logits = outputs.logits
+                predicted_class_id = logits.argmax(-1).item()
+                confidence = torch.softmax(logits, dim=-1).max().item()
+            
+            # 獲取食物名稱
+            food_name = model.config.id2label[predicted_class_id]
+            
+            log_analysis_step("識別完成", f"識別結果: {food_name} (信心度: {confidence:.1%})")
+            
+            # 查詢營養資訊
+            nutrition_data = lookup_nutrition_data(food_name)
+            
+            return {
+                "food_name": food_name,
+                "confidence": confidence * 100,
+                "weight": 150,  # 模擬重量估算
+                "nutrition": nutrition_data.get("nutrition", {}),
+                "analysis_steps": [
+                    "Hugging Face Food101 模型載入完成",
+                    "圖片下載和預處理完成",
+                    f"食物識別: {food_name}",
+                    "營養資訊查詢完成"
+                ],
+                "mode": "real_hf_model"
+            }
+            
+        except Exception as model_error:
+            log_analysis_step("模型載入", f"Hugging Face 模型載入失敗: {str(model_error)}")
+            log_analysis_step("模式切換", "切換到模擬模式")
+            return simulate_ai_analysis(image_url)
         
     except Exception as e:
         log_analysis_step("模型載入", f"模型載入失敗: {str(e)}")

requirements.txt

@@ -4,8 +4,15 @@ uvicorn[standard]>=0.27.0
 gradio>=4.44.0
 pydantic>=2.5.0
 
+# Hugging Face models
+transformers>=4.35.0
+torch>=2.0.0
+torchvision>=0.15.0
+diffusers>=0.24.0
+
 # Image processing
 Pillow>=10.3.0
+opencv-python>=4.8.0
 
 # HTTP requests
 requests>=2.31.0
@@ -13,6 +20,7 @@ httpx>=0.25.0
 
 # Data processing
 numpy>=1.24.0
+pandas>=2.0.0
 
 # CORS support
 python-multipart>=0.0.6