Update app.py

app.py

@@ -13,6 +13,9 @@ clip_processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
 blip_processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
 blip_model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
 
+# 定义CLIP特征的名称（假设的特征名称，您可以根据需要调整）
+CLIP_FEATURE_NAMES = [f"Dimension {i}" for i in range(512)]
+
 # 图像处理函数
 def compute_difference_images(img_a, img_b):
     def extract_sketch(image):
@@ -37,10 +40,10 @@ def compute_difference_images(img_a, img_b):
     }
 
 # 保存图像到文件
-def save_images(images):
+def save_images(images, prefix):
     paths = []
     for key, img in images.items():
-        path = f"{key}.png"
+        path = f"{prefix}_{key}.png"
         img.save(path)
         paths.append((path, key.replace("_", " ").capitalize()))
     return paths
@@ -52,23 +55,29 @@ def generate_detailed_caption(image):
     return blip_processor.decode(caption[0], skip_special_tokens=True)
 
 # 特征差异可视化
-def plot_feature_differences(latent_diff):
+def plot_feature_differences(latent_diff, prefix):
     diff_magnitude = [abs(x) for x in latent_diff[0]]
     indices = range(len(diff_magnitude))
+    top_indices = np.argsort(diff_magnitude)[-10:][::-1]  # Top 10 differences
 
     plt.figure(figsize=(8, 4))
     plt.bar(indices, diff_magnitude, alpha=0.7)
     plt.xlabel("Feature Index (Latent Dimension)")
     plt.ylabel("Magnitude of Difference")
     plt.title("Feature Differences (Bar Chart)")
-    bar_chart_path = "bar_chart.png"
+    bar_chart_path = f"{prefix}_bar_chart.png"
     plt.savefig(bar_chart_path)
     plt.close()
 
     plt.figure(figsize=(6, 6))
-    plt.pie(diff_magnitude[:10], labels=[f"Feature {i}" for i in range(10)], autopct="%1.1f%%", startangle=140)
+    plt.pie(
+        [diff_magnitude[i] for i in top_indices],
+        labels=[CLIP_FEATURE_NAMES[i] for i in top_indices],
+        autopct="%1.1f%%",
+        startangle=140
+    )
     plt.title("Top 10 Feature Differences (Pie Chart)")
-    pie_chart_path = "pie_chart.png"
+    pie_chart_path = f"{prefix}_pie_chart.png"
     plt.savefig(pie_chart_path)
     plt.close()
 
@@ -88,13 +97,12 @@ def generate_text_analysis(api_key, api_type, caption_a, caption_b):
             {"role": "user", "content": f"图片A的描述为：{caption_a}。图片B的描述为：{caption_b}。\n请对两张图片的内容和潜在特征区别进行详细分析，并输出一个简洁但富有条理的总结。"}
         ]
     )
-    # 修复: 正确访问返回值
     return response.choices[0].message.content.strip()
 
 # 分析函数
-def analyze_images(img_a, img_b, api_key, api_type):
+def analyze_images(img_a, img_b, api_key, api_type, prefix):
     images_diff = compute_difference_images(img_a, img_b)
-    saved_images = save_images(images_diff)
+    saved_images = save_images(images_diff, prefix)
 
     caption_a = generate_detailed_caption(img_a)
     caption_b = generate_detailed_caption(img_b)
@@ -107,7 +115,7 @@ def analyze_images(img_a, img_b, api_key, api_type):
 
     latent_diff = np.abs(features_a - features_b).tolist()
 
-    bar_chart, pie_chart = plot_feature_differences(latent_diff)
+    bar_chart, pie_chart = plot_feature_differences(latent_diff, prefix)
     text_analysis = generate_text_analysis(api_key, api_type, caption_a, caption_b)
 
     return {
@@ -125,7 +133,8 @@ def batch_analyze(images_a, images_b, api_key, api_type):
 
     results = []
     for i in range(num_pairs):
-        result = analyze_images(images_a[i], images_b[i], api_key, api_type)
+        prefix = f"comparison_{i+1}"
+        result = analyze_images(images_a[i], images_b[i], api_key, api_type, prefix)
         results.append({
             "pair": (f"Image A-{i+1}", f"Image B-{i+1}"),
             **result