Ai Image Detect Distilled

🚀 AI Detection Model

An AI detection model for image classification, capable of distinguishing between real and AI - generated images.

🚀 Quick Start

This README provides detailed information about the AI Image Detect Distilled model, including its architecture, training process, data sources, performance, and future directions.

✨ Features

• Multi - model Distillation: Combines the features learned from three separate models into a small ViT model for efficient detection.
• Diverse Data Sources: Utilizes multiple datasets to ensure the similarity between real and AI - generated images.
• Good Performance: Achieves high accuracy on both validation and real - world datasets, outperforming other popular models.

📦 Installation

No installation steps are provided in the original document.

💻 Usage Examples

No code examples are provided in the original document.

📚 Documentation

Model Architecture and Training

Three separate models were initially trained:

1. Midjourney vs. Real Images
2. Stable Diffusion vs. Real Images
3. Stable Diffusion Fine - tunings vs. Real Images

The data preparation process was as follows:

• Used Google's Open Image Dataset for real images
• Described real images using BLIP (Bootstrapping Language - Image Pre - training)
• Generated Stable Diffusion images using BLIP descriptions
• Found similar Midjourney images based on BLIP descriptions

This approach ensured that real and AI - generated images were as similar as possible, differing only in their origin.

The three models were then distilled into a small ViT model with 11.8 Million Parameters, combining their learned features for more efficient detection.

Data Sources

• Google's Open Image Dataset: link
• Ivan Sivkov's Midjourney Dataset: link
• TANREI(NAMA)'s Stable Diffusion Prompts Dataset: link

Performance

• Validation Set: 74% accuracy. It was held out from the training data to assess generalization.
• Custom Real - World Set: 72% accuracy. Composed of self - captured images and online - sourced images, it is designed to be more representative of internet - based images.
• Comparative Analysis: Outperformed other popular AI detection models by 5 percentage points on both sets. Other models achieved 89% and 79% on the validation and real - world sets respectively.

Key Insights

1. Strong generalization on validation data (75% accuracy).
2. Good adaptability to diverse, real - world images (72% accuracy).
3. Consistent outperformance of other popular models.
4. A 10 - point accuracy drop from the validation to the real - world set indicates room for improvement.
5. Comprehensive training on multiple AI generation techniques contributes to model versatility.
6. Focus on subtle differences in image generation rather than content disparities.

Future Directions

• Expand the dataset with more diverse, real - world examples to bridge the performance gap.
• Improve generalization to internet - sourced images.
• Conduct error analysis on misclassified samples to identify patterns.
• Integrate new AI image generation techniques as they emerge.
• Consider fine - tuning for specific domains where detection accuracy is critical.

🔧 Technical Details

The model architecture involves distilling three separate models into a small ViT model with 11.8 Million Parameters. The data preparation process carefully aligns real and AI - generated images to ensure similarity in appearance.

📄 License

This project is licensed under the MIT license.