Update app.py

app.py

@@ -514,7 +514,7 @@ def analyze_clothing_details(image):
                 features["clothing_angle_uniformity"] = float(max_count / max(total_count, 1))
         
         # 分析纹理的一致性
- # 将图像分成小块，计算每个块的纹理特征
+        # 将图像分成小块，计算每个块的纹理特征
         block_size = 32
         h, w = gray.shape
         texture_variations = []
@@ -534,7 +534,7 @@ def analyze_clothing_details(image):
             
             # 计算纹理变化的熵 - 真实衣物纹理变化的熵更高
             features["clothing_texture_entropy"] = float(stats.entropy(texture_variations + 1e-10))
-# 褶皱分析 - 使用形态学操作提取可能的褶皱
+ # 褶皱分析 - 使用形态学操作提取可能的褶皱
         _, thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
         kernel = np.ones((3,3), np.uint8)
         opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
@@ -654,15 +654,39 @@ def analyze_extremities(image):
     return features
 
 def check_ai_specific_features(image_features):
-    """检查AI生成图像的典型特征"""
+    """检查AI生成图像的典型特征，增强对最新AI模型的检测能力"""
     ai_score = 0
     ai_signs = []
     
-    # 检查对称性 - AI生成图像通常对称性高，但权重降低
+    # 增加对微观纹理的分析权重
+    if "lbp_entropy" in image_features:
+        if image_features["lbp_entropy"] < 2.0:
+            ai_score += 0.3  # 提高权重
+            ai_signs.append("微观纹理熵极低，典型AI生成特征")
+        elif image_features["lbp_entropy"] < 3.0:
+            ai_score += 0.2
+            ai_signs.append("微观纹理熵异常低")
+    
+    # 增加对频率分布各向异性的分析权重
+    if "freq_anisotropy" in image_features:
+        if image_features["freq_anisotropy"] < 0.05:
+            ai_score += 0.3  # 提高权重
+            ai_signs.append("频率分布各向异性极低，典型AI生成特征")
+        elif image_features["freq_anisotropy"] < 0.5:
+            ai_score += 0.2
+            ai_signs.append("频率分布各向异性异常低")
+    
+    # 增加对细节一致性的分析
+    if "detail_scale_consistency" in image_features and "detail_spatial_std" in image_features:
+        if image_features["detail_scale_consistency"] < 0.2 and image_features["detail_spatial_std"] < 5:
+            ai_score += 0.3
+            ai_signs.append("细节一致性异常，典型AI生成特征")
+    
+    # 检查对称性 - AI生成图像通常对称性高
     if "horizontal_symmetry" in image_features and "vertical_symmetry" in image_features:
         avg_symmetry = (image_features["horizontal_symmetry"] + image_features["vertical_symmetry"]) / 2
-        if avg_symmetry > 0.8:  # 提高阈值
-            ai_score += 0.15  # 降低权重
+        if avg_symmetry > 0.8:
+            ai_score += 0.15
             ai_signs.append("图像对称性异常高")
         elif avg_symmetry > 0.7:
             ai_score += 0.1
@@ -705,26 +729,6 @@ def check_ai_specific_features(image_features):
             ai_score += 0.1
             ai_signs.append("频率分布不自然")
     
-    # 检查频率各向异性 - AI生成图像通常频率分布各向异性较低
-    if "freq_anisotropy" in image_features and image_features["freq_anisotropy"] < 0.5:
-        ai_score += 0.15
-        ai_signs.append("频率分布各向异性异常低")
-    
-    # 检查局部颜色变化 - 真实照片通常有更多局部颜色变化
-    if "local_color_variation" in image_features and image_features["local_color_variation"] < 5:
-        ai_score += 0.15
-        ai_signs.append("局部颜色变化异常少")
-    
-    # 检查边缘变化 - 真实照片的边缘通常更自然
-    if "edge_variance" in image_features and image_features["edge_variance"] < 100:
-        ai_score += 0.15
-        ai_signs.append("边缘变化异常均匀")
-    
-    # 检查边缘方向熵 - 真实照片的边缘方向熵通常更高
-    if "edge_direction_entropy" in image_features and image_features["edge_direction_entropy"] < 1.5:
-        ai_score += 0.15
-        ai_signs.append("边缘方向熵异常低")
-    
     # 检查噪声频谱 - 真实照片的噪声频谱更自然
     if "noise_spectrum_std" in image_features and image_features["noise_spectrum_std"] < 1000:
         ai_score += 0.15
@@ -817,10 +821,13 @@ def check_ai_specific_features(image_features):
         ai_score += 0.15
         ai_signs.append("手部曲率变化异常均匀")
     
-    # 微观纹理分析 - AI生成图像的微观纹理通常不够自然
-    if "lbp_entropy" in image_features and image_features["lbp_entropy"] < 3.0:
-        ai_score += 0.2
-        ai_signs.append("微观纹理熵异常低")
+    # 特别关注最新AI模型的特征组合
+    # 当多个特征同时出现时，这是强有力的AI生成证据
+    ai_feature_count = len(ai_signs)
+    if ai_feature_count >= 5:  # 如果检测到多个AI特征
+        ai_score = max(ai_score, 0.9)  # 确保AI分数很高
+    elif ai_feature_count >= 3:
+        ai_score = max(ai_score, 0.7)
     
     return min(ai_score, 1.0), ai_signs
 
@@ -954,7 +961,7 @@ def detect_photoshop_signs(image_features):
     return min(ps_score, 1.0), ps_signs
 
 def get_detailed_analysis(ai_probability, ps_score, beauty_score, ps_signs, ai_signs, beauty_signs, valid_models_count):
-    """提供更详细的分析结果"""
+    """提供更详细的分析结果，使用二级分类框架"""
     
     # 根据有效模型数量调整置信度描述
     confidence_prefix = ""
@@ -965,46 +972,30 @@ def get_detailed_analysis(ai_probability, ps_score, beauty_score, ps_signs, ai_s
     elif valid_models_count == 1:
         confidence_prefix = "中等置信度："
     
-    # 调整后的阈值判断，考虑美颜因素和衣物细节
-    if ai_probability > 0.7:  # 高AI概率
-        category = confidence_prefix + "高概率AI生成"
+    # 第一级分类：AI生成 vs 真人照片
+    if ai_probability > 0.6:  # 降低AI判定阈值，提高AI检出率
+        category = confidence_prefix + "AI生成图像"
         description = "图像很可能是由AI完全生成，几乎没有真人照片的特征。"
-    elif ai_probability > 0.5:  # 中等AI概率
-        if beauty_score > 0.6:  # 高美颜分数
-            category = confidence_prefix + "可能是重度美颜的真人照片"
-            description = "图像可能是真人照片经过重度美颜处理，也可能是AI生成图像。"
-        elif ps_score > 0.5:  # 高PS分数
-            category = confidence_prefix + "中等概率AI生成，高概率PS修图"
-            description = "图像可能是真人照片经过大量后期处理，或是AI生成后经过修饰的图像。"
-        else:
-            category = confidence_prefix + "中等概率AI生成"
-            description = "图像有较多AI生成的特征，但也保留了一些真实照片的特点。"
-    elif ai_probability > 0.3:  # 低AI概率
-        if beauty_score > 0.5:  # 中高美颜分数
-            category = confidence_prefix + "很可能是美颜处理的真人照片"
-            description = "图像很可能是真人照片经过美颜处理，美颜痕迹明显。"
-        elif ps_score > 0.5:  # 高PS分数
-            category = confidence_prefix + "低概率AI生成，高概率PS修图"
-            description = "图像更可能是真人照片经过大量后期处理，PS痕迹明显。"
-        else:
-            category = confidence_prefix + "低概率AI生成"
-            description = "图像更可能是真人照片，但有一些AI生成或修饰的特征。"
-    else:  # 很低AI概率
-        if beauty_score > 0.6:
-            category = confidence_prefix + "真人照片，重度美颜处理"
-            description = "图像基本是真人照片，但经过了重度美颜处理。"
-        elif ps_score > 0.6:
-            category = confidence_prefix + "真人照片，重度PS修图"
-            description = "图像基本是真人照片，但经过了大量后期处理，修饰痕迹明显。"
-        elif ps_score > 0.3 or beauty_score > 0.3:
-            category = confidence_prefix + "真人照片，中度修图或美颜"
-            description = "图像是真人照片，有明显的后期处理或美颜痕迹。"
-        elif ps_score > 0.1 or beauty_score > 0.1:
-            category = confidence_prefix + "真人照片，轻度修图或美颜"
-            description = "图像是真人照片，有少量后期处理或美颜。"
+        main_category = "AI生成"
+    else:
+        # 第二级分类：真实素人 vs 修图痕迹明显
+        combined_edit_score = max(ps_score, beauty_score)  # 取PS和美颜中的较高分
+        
+        if combined_edit_score > 0.5:
+            category = confidence_prefix + "真人照片，修图痕迹明显"
+            description = "图像基本是真人照片，但经过了明显的后期处理或美颜，修饰痕迹明显。"
+            main_category = "真人照片-修图明显"
         else:
-            category = confidence_prefix + "高概率真人照片，几乎无修图"
-            description = "图像几乎可以确定是未经大量处理的真人照片。"
+            category = confidence_prefix + "真实素人照片"
+            description = "图像很可能是未经大量处理的真人照片，保留了自然的细节和特征。"
+            main_category = "真人照片-素人"
+    
+    # 处理边界情况 - 当AI概率和修图分数都很高时
+    if ai_probability > 0.45 and combined_edit_score > 0.7:
+        # 这是一个边界情况，可能是高度修图的真人照片，也可能是AI生成的
+        category = confidence_prefix + "真人照片，修图痕迹明显（也可能是AI生成）"
+        description = "图像可能是真人照片经过大量后期处理，也可能是AI生成图像。由于现代AI技术与高度修图效果相似，难以完全区分。"
+        main_category = "真人照片-修图明显"
     
     # 添加具体的PS痕迹描述
     if ps_signs:
@@ -1024,7 +1015,7 @@ def get_detailed_analysis(ai_probability, ps_score, beauty_score, ps_signs, ai_s
     else:
         beauty_details = "未检测到明显的美颜特征。"
     
-    return category, description, ps_details, ai_details, beauty_details
+    return category, description, ps_details, ai_details, beauty_details, main_category
 
 def detect_ai_image(image):
     """主检测函数"""
@@ -1073,8 +1064,7 @@ def detect_ai_image(image):
         final_ai_probability = weighted_ai_probability / sum(m["weight"] for k, m in models.items() if m["processor"] is not None and m["model"] is not None)
     else:
         return {"error": "所有模型加载失败"}
-    
-    # 分析图像特征
+ # 分析图像特征
     image_features = analyze_image_features(image)
     
     # 检查AI特定特征
@@ -1089,13 +1079,41 @@ def detect_ai_image(image):
     # 应用特征权重调整AI概率
     adjusted_probability = final_ai_probability
     
-    # 如果AI特征分数高，提高AI概率
-    if ai_feature_score > 0.7:
+    # 提高AI特征分数的权重
+    if ai_feature_score > 0.8:  # 当AI特征非常明显时
+        adjusted_probability = max(adjusted_probability, 0.8)  # 大幅提高AI概率
+    elif ai_feature_score > 0.6:
         adjusted_probability = max(adjusted_probability, 0.7)
-    elif ai_feature_score > 0.5:
+    elif ai_feature_score > 0.4:
         adjusted_probability = max(adjusted_probability, 0.6)
-    elif ai_feature_score > 0.3:
-        adjusted_probability = max(adjusted_probability, 0.5)
+    
+    # 特别关注关键AI特征
+    key_ai_features_count = 0
+    
+    # 检查微观纹理熵 - 这是AI生成的强力指标
+    if "lbp_entropy" in image_features and image_features["lbp_entropy"] < 2.5:
+        key_ai_features_count += 1
+        adjusted_probability += 0.1
+    
+    # 检查频率分布各向异性 - 这是AI生成的强力指标
+    if "freq_anisotropy" in image_features and image_features["freq_anisotropy"] < 0.1:
+        key_ai_features_count += 1
+        adjusted_probability += 0.1
+    
+    # 检查细节空间分布 - 这是AI生成的强力指标
+    if "detail_spatial_std" in image_features and image_features["detail_spatial_std"] < 5:
+        key_ai_features_count += 1
+        adjusted_probability += 0.1
+    
+    # 如果多个关键AI特征同时存在，这是强有力的AI生成证据
+    if key_ai_features_count >= 2:
+        adjusted_probability = max(adjusted_probability, 0.7)
+    
+    # 降低美颜特征对AI判断的影响
+    # 即使美颜分数高，如果AI特征也明显，仍应判定为AI生成
+    if beauty_score > 0.6 and ai_feature_score > 0.7:
+        # 不再降低AI概率，而是保持较高的AI概率
+        pass
     
     # 如果检测到衣物或手部异常，大幅提高AI概率
     if "clothing_angle_uniformity" in image_features and image_features["clothing_angle_uniformity"] > 0.3:
@@ -1104,18 +1122,6 @@ def detect_ai_image(image):
     if "extremity_perimeter_area_ratio" in image_features and image_features["extremity_perimeter_area_ratio"] < 0.05:
         adjusted_probability = max(adjusted_probability, 0.7)
     
-    # 如果检测到微观纹理异常，提高AI概率
-    if "lbp_entropy" in image_features and image_features["lbp_entropy"] < 3.0:
-        adjusted_probability = max(adjusted_probability, 0.6)
-    
-    # 如果检测到频率分布异常，提高AI概率
-    if "freq_anisotropy" in image_features and image_features["freq_anisotropy"] < 0.5:
-        adjusted_probability = max(adjusted_probability, 0.6)
-    
-    # 如果美颜分数高但AI特征分数不高，降低AI概率
-    if beauty_score > 0.6 and ai_feature_score < 0.5:
-        adjusted_probability = min(adjusted_probability, 0.5)
-    
     # 确保概率在0-1范围内
     adjusted_probability = min(1.0, max(0.0, adjusted_probability))
     
@@ -1127,7 +1133,7 @@ def detect_ai_image(image):
             adjusted_probability = (adjusted_probability + ai_detector_prob * 2) / 3
     
     # 获取详细分析
-    category, description, ps_details, ai_details, beauty_details = get_detailed_analysis(
+    category, description, ps_details, ai_details, beauty_details, main_category = get_detailed_analysis(
         adjusted_probability, ps_score, beauty_score, ps_signs, ai_signs, beauty_signs, valid_models
     )
     
@@ -1139,6 +1145,7 @@ def detect_ai_image(image):
         "beauty_score": beauty_score,
         "ai_feature_score": ai_feature_score,
         "category": category,
+        "main_category": main_category,  # 添加主要分类
         "description": description,
         "ps_details": ps_details,
         "ai_details": ai_details,
@@ -1153,9 +1160,12 @@ def detect_ai_image(image):
 iface = gr.Interface(
     fn=detect_ai_image,
     inputs=gr.Image(type="pil"),
-    outputs=gr.JSON(),
+    outputs=[
+        gr.JSON(label="详细分析结果"),
+        gr.Label(label="主要分类", num_top_classes=1)
+    ],
     title="增强型AI图像检测API",
-    description="多模型集成检测图像是否由AI生成，同时分析PS修图和美颜痕迹，特别关注通用细节特征",
+    description="多模型集成检测图像是否由AI生成或真人照片（素人/修图）",
     examples=None,
     allow_flagging="never"
 )