| Here's a `README.md` for your project: | |
| ```markdown | |
| # Video Manipulation Detection Using AI | |
| This project uses an AI-powered video analysis pipeline to detect potential video manipulations such as deepfakes and other visual tampering. It leverages the **DEtection TRansformer (DETR)** model for object detection and optical flow analysis to identify motion inconsistencies. The application is built using **Streamlit**, and the model processes videos to generate detailed reports on their authenticity. | |
| ## Features | |
| - **Metadata Extraction**: Extracts and displays metadata from uploaded videos. | |
| - **Frame Extraction**: Extracts frames from videos at specified intervals for analysis. | |
| - **Object Detection**: Analyzes each frame to detect objects using the DETR model. | |
| - **Optical Flow Calculation**: Computes motion patterns to detect inconsistencies. | |
| - **Report Generation**: Compiles visual and motion analysis results into a comprehensive JSON report. | |
| - **Manipulation Detection**: Uses the report to assess the likelihood of video manipulation. | |
| ## Technologies Used | |
| - **Streamlit**: Front-end web framework for creating the user interface. | |
| - **PyTorch**: Machine learning framework used for object detection (DETR). | |
| - **OpenCV**: Library for video and image processing (frame extraction, optical flow). | |
| - **Transformers**: Pre-trained DETR model for object detection. | |
| - **FFmpeg**: Tool for extracting metadata from videos and processing video files. | |
| ## Installation | |
| 1. Clone the repository: | |
| ```bash | |
| git clone https://github.com/your-username/video-manipulation-detection.git | |
| cd video-manipulation-detection | |
| ``` | |
| 2. Install dependencies: | |
| ```bash | |
| pip install -r requirements.txt | |
| ``` | |
| 3. Run the Streamlit app: | |
| ```bash | |
| streamlit run app.py | |
| ``` | |
| ## How to Use | |
| 1. **Upload a Video**: Go to the "Analyze Video" section and upload a `.mp4` or `.mov` video file. | |
| 2. **Start Analysis**: Click the **"Analyze Video"** button to start processing. | |
| 3. **View Results**: After analysis, the app will display a score and details about the potential manipulation of the video. | |
| ## Report Details | |
| The report includes the following information: | |
| - **Frame Analysis**: Number of detections, maximum and average confidence of the object detection model. | |
| - **Motion Analysis**: Calculated optical flow values to detect motion inconsistencies. | |
| - **Summary Stats**: Overall detection confidence, optical flow statistics, and manipulation detection score. | |
| ## License | |
| This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details. | |
| ## Acknowledgments | |
| - [Streamlit](https://streamlit.io/) | |
| - [PyTorch](https://pytorch.org/) | |
| - [Hugging Face Transformers](https://huggingface.co/) | |
| - [OpenCV](https://opencv.org/) | |
| - [FFmpeg](https://ffmpeg.org/) | |
| - This project is built as part of a research and development effort to analyze and detect video manipulations. | |
| ## Contact | |
| - **Ayush Agarwal** | |
| AIML Enthusiast, VIT Bhopal University | |
| 📧 [[email protected]](mailto:[email protected]) | |
| 🔗 [LinkedIn](https://www.linkedin.com/in/ayush20039939) | |
| 🐙 [GitHub](https://github.com/your-github) | |
| ``` | |
| ### Key Sections: | |
| 1. **Project Overview**: Describes the purpose and features of the app. | |
| 2. **Technologies Used**: Lists the libraries and tools used in the project. | |
| 3. **Installation**: Step-by-step guide on how to clone, install dependencies, and run the app. | |
| 4. **How to Use**: Instructions on how users can interact with the app. | |
| 5. **Report Details**: Describes the information generated in the analysis report. | |
| 6. **License & Acknowledgments**: Includes the MIT License and credits to the libraries and tools used. | |
| This will help users understand the project and how to deploy and use it. Feel free to customize it further! | |