File size: 7,362 Bytes
9ea88f7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
"""
Prediction script combining DINOv2 classifier and Qwen2-VL reasoner
Outputs predictions.json in required format
"""

import torch
import torch.nn as nn
from torchvision import transforms
from transformers import (
    AutoImageProcessor, 
    Dinov2Model,
    Qwen3VLForConditionalGeneration,
    AutoProcessor
)
from peft import PeftModel
from PIL import Image
import json
import os
from pathlib import Path
from tqdm import tqdm
from qwen_vl_utils import process_vision_info

class DINOv2Classifier(nn.Module):
    def __init__(self, num_classes=3):
        super().__init__()
        self.dinov2 = Dinov2Model.from_pretrained("facebook/dinov2-base")
        
        # Classification head
        self.classifier = nn.Sequential(
            nn.Linear(768, 512),
            nn.ReLU(),
            nn.Dropout(0.3),
            nn.Linear(512, 256),
            nn.ReLU(),
            nn.Dropout(0.3),
            nn.Linear(256, num_classes)
        )
        
    def forward(self, pixel_values):
        outputs = self.dinov2(pixel_values)
        cls_token = outputs.last_hidden_state[:, 0]
        logits = self.classifier(cls_token)
        return logits

class GenAIDetector:
    def __init__(self, classifier_path):
        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
        print(f"Using device: {self.device}")
        
        # Load DINOv2 classifier
        print("Loading classifier...")
        self.classifier = DINOv2Classifier(num_classes=3).to(self.device)
        checkpoint = torch.load(classifier_path, map_location=self.device)
        self.classifier.load_state_dict(checkpoint['model_state_dict'])
        self.classifier.eval()
        
        self.image_processor = AutoImageProcessor.from_pretrained("facebook/dinov2-base")
        
        # Load VLM
        print("Loading VLM reasoner...")
        base_model = Qwen3VLForConditionalGeneration.from_pretrained(
            "Qwen/Qwen3-VL-8B-Instruct",
            torch_dtype="auto",
            device_map="auto"
        )
        self.vlm = base_model
        self.vlm_processor = AutoProcessor.from_pretrained("Qwen/Qwen3-VL-8B-Instruct")
        self.vlm.eval()
        
        self.class_names = ['real', 'manipulated', 'fake']
        self.manipulation_types = {
            'real': 'none',
            'manipulated': 'inpainting',
            'fake': 'full_synthesis'
        }
    
    def classify_image(self, image_path):
        """Classify image and get confidence scores"""
        image = Image.open(image_path).convert('RGB')
        inputs = self.image_processor(images=image, return_tensors="pt")
        pixel_values = inputs['pixel_values'].to(self.device)
        
        with torch.no_grad():
            logits = self.classifier(pixel_values)
            probs = torch.softmax(logits, dim=1)
            pred_class = torch.argmax(probs, dim=1).item()
            confidence = probs[0].cpu().numpy()
        
        return pred_class, confidence
    
    def generate_reasoning(self, image_path, predicted_class):
        """Generate reasoning using VLM"""
        class_name = self.class_names[predicted_class]
        
        # Prepare prompt
        prompt = f"The given image has been flagged as {class_name}. Explain in 2-3 sentences why that might be. Focus on specific features which indicated this."
        
        messages = [
            {
                "role": "user",
                "content": [
                    {"type": "image", "image": image_path},
                    {"type": "text", "text": prompt}
                ]
            }
        ]
        
        # Apply chat template
        text = self.vlm_processor.apply_chat_template(
            messages,
            tokenize=False,
            add_generation_prompt=True
        )
        
        # Process inputs
        image_inputs, video_inputs = process_vision_info(messages)
        inputs = self.vlm_processor(
            text=[text],
            images=image_inputs,
            videos=video_inputs,
            padding=True,
            return_tensors="pt"
        )
        inputs = inputs.to(self.device)
        
        # Generate
        with torch.no_grad():
            output_ids = self.vlm.generate(
                **inputs,
                max_new_tokens=150,
                temperature=0.7,
                do_sample=True
            )
        
        # Decode
        generated_text = self.vlm_processor.batch_decode(
            output_ids,
            skip_special_tokens=True,
            clean_up_tokenization_spaces=False
        )[0]
        
        # Extract assistant response
        if "assistant" in generated_text.lower():
            reasoning = generated_text.split("assistant")[-1].strip()
        else:
            reasoning = generated_text.strip()
        
        return reasoning
    
    def predict(self, image_path):
        """Full prediction pipeline"""
        # Classify
        pred_class, confidence = self.classify_image(image_path)
        
        # Get authenticity score (confidence that it's real, i.e., confidence[0])
        authenticity_score = float(1.0 - confidence[0])  # Higher score = more manipulated
        
        # Get manipulation type
        class_name = self.class_names[pred_class]
        manipulation_type = self.manipulation_types[class_name]
        
        # Generate reasoning
        reasoning = self.generate_reasoning(image_path, pred_class)
        
        return {
            'authenticity_score': round(authenticity_score, 2),
            'manipulation_type': manipulation_type,
            'vlm_reasoning': reasoning
        }

def main(image_dir, classifier_path, output_file):
    """Main prediction function"""
    
    # Initialize detector
    detector = GenAIDetector(classifier_path)
    
    # Get all images
    image_extensions = ['.jpg', '.jpeg', '.png']
    image_files = []
    for ext in image_extensions:
        image_files.extend(Path(image_dir).glob(f'*{ext}'))
        image_files.extend(Path(image_dir).glob(f'*{ext.upper()}'))
    
    print(f"Found {len(image_files)} images")
    
    # Process images
    predictions = []
    for image_path in tqdm(image_files, desc="Processing images"):
        try:
            result = detector.predict(str(image_path))
            result['image_name'] = image_path.name
            predictions.append(result)
        except Exception as e:
            print(f"Error processing {image_path.name}: {str(e)}")
            continue
    
    # Save predictions
    with open(output_file, 'w') as f:
        json.dump(predictions, f, indent=2)
    
    print(f"\n✓ Processed {len(predictions)} images")
    print(f"✓ Saved predictions to {output_file}")

if __name__ == "__main__":
    import argparse
    
    parser = argparse.ArgumentParser()
    parser.add_argument('--image_dir', type=str, default='./test_images',
                        help='Directory containing images to predict')
    parser.add_argument('--classifier_path', type=str, default='best_model.pth',
                        help='Path to trained DINOv2 checkpoint (.pth file)')
    parser.add_argument('--output_file', type=str, default='predictions.json',
                        help='Output JSON file')
    
    args = parser.parse_args()
    
    main(args.image_dir, args.classifier_path, args.output_file)