Detect_AI_vs_Human_Generated_Image

A deep learning-based model to distinguish between AI-generated and human-created images. This repository includes data preprocessing, model training (ConVitX with ConvNeXt Large and Swin Transformer), evaluation using F1-score, and real-time deployment. Built for the Kaggle competition "Can You Tell the Difference?"

Table of Contents

• Introduction
• Dataset
• Model Architecture
• Training
• Evaluation
• Deployment
• Usage
• Files
• Contributing
• License
• Contact

Introduction

The goal of this project is to develop a deep learning model capable of distinguishing between images generated by AI and those created by humans. This is achieved using a combination of ConvNeXt Large and Swin Transformer architectures, named ConVitX.

Dataset

The dataset used for this project is provided by the Kaggle competition "Can You Tell the Difference?". The dataset includes a variety of images, both AI-generated and human-created.

Model Architecture

The model architecture consists of:

• ConVitX: A custom model constructed by combining ConvNeXt Large and Swin Transformer architectures.
• ConvNeXt Large: A convolutional network architecture designed for high performance on image classification tasks.
• Swin Transformer: A transformer architecture with shifted windows for efficient image processing.

Training

The model is trained using the following steps:

1. Data preprocessing
2. Model training with ConVitX
3. Hyperparameter tuning

Evaluation

The model is evaluated using the F1-score metric to ensure a balanced measure of precision and recall.

I secured 25th position on the public leaderboard

Deployment

The trained model is deployed for real-time image classification, allowing users to input images and receive predictions on whether they are AI-generated or human-created.

Usage

To use the model for image classification, follow these steps:

1. Clone the repository: git clone https://github.com/SheemaMasood381/Detect_AI_vs_Human_Generated_Image.git
2. Install the required dependencies: pip install -r requirements.txt
3. Run the classification script: python classify_image.py --image_path path_to_image

Files

• README.md: Project overview and instructions
• app.py: Main application script for deployment
• inference.py: Script for making inferences with the trained model
• lb.PNG: Leaderboard snapshot showing 28th position
• load_model.py: Script for loading the trained model
• model.py: Model architecture definition
• notebooks/: Directory containing Jupyter notebooks for experimentation
• requirements.txt: List of dependencies required for the project

Contributing

Contributions are welcome! Please follow these steps to contribute:

1. Fork the repository
2. Create a new branch: git checkout -b feature-branch
3. Make your changes
4. Commit your changes: git commit -m 'Add new feature'
5. Push to the branch: git push origin feature-branch
6. Create a pull request

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

Feel free to reach out to me on LinkedIn for any queries or collaborations.