Update app.py

app.py

@@ -114,42 +114,18 @@ def infer(image_pil: Image.Image,
 
     transform = build_transform(is_training=False, interpolation=interpolation)
     input_tensor = transform(image_pil).unsqueeze(0).to(device)
-
-    # (1) 分类预测
+    
     logits = model(input_tensor)
     probs  = F.softmax(logits, dim=1)[0]
     confidences = {class_names[i]: float(probs[i]) for i in range(NUM_CLASSES)}
 
-    if not show_cam:
-        return confidences, None
-
-    # (2) Grad-CAM（多层级融合 → 均值）
-    rgb_img = np.asarray(image_pil).astype(np.float32) / 255.0
-    targets = [ClassifierOutputTarget(1)]      # 固定看 AI 分类(索引 = 1)
-
-    with GradCAM(model=model,
-                 target_layers=target_layers,
-                 reshape_transform=reshape_transform_swin) as cam:
-
-        all_cams = cam(input_tensor, targets=targets,
-                       aug_smooth=True, eigen_smooth=True)  # shape: [L, H', W']
-
-    grayscale_cam = np.mean(all_cams, axis=0)               # L 层取平均融合
-    heatmap = Image.fromarray((grayscale_cam * 255).astype(np.uint8)) \
-                    .resize(image_pil.size, Image.Resampling.BILINEAR)
-
-    cam_img = show_cam_on_image(rgb_img,
-                                np.asarray(heatmap).astype(np.float32) / 255.,
-                                use_rgb=True)
-    cam_pil = Image.fromarray(cam_img)
-
-    return confidences, cam_pil
+    return confidences
 
 # ---------------------------------------------------------------------------
 # 7. Gradio UI
 def launch_app():
     with gr.Blocks() as demo:
-        gr.Markdown("# 🖼️ AI vs. Non-AI Image Classifier  (Swin-Large + Grad-CAM)")
+        gr.Markdown("# 🖼️ AI vs. Non-AI Image Classifier  (with Swin-Large)")
 
         run_btn = gr.Button("🚀 Run")
         
@@ -163,8 +139,8 @@ def launch_app():
             in_img = gr.Image(type="pil", label="Upload an Image")
             out_lbl = gr.Label(num_top_classes=2, label="Predictions")
 
-        def _run(img, inter, cam_flag):
-            return infer(img, interpolation=inter, show_cam=False)
+        def _run(img, inter):
+            return infer(img, interpolation=inter)
 
         run_btn.click(
             _run,