metadata

license: afl-3.0
title: AI - Video Manipulation Detectio
sdk: streamlit
emoji: 🏆
colorFrom: yellow
colorTo: blue
short_description: 'This project is an AI-powered tool designed to detect video '
sdk_version: 1.47.0

Video Manipulation Detection - Streamlit App

Overview

This project is an AI-powered tool designed to detect video manipulations like deepfakes using object detection and optical flow analysis. It provides a web interface where users can upload videos, and the tool analyzes the video for potential manipulations, giving a score to indicate the likelihood of manipulation.

Configuration

title: Video Manipulation Detection
emoji: 🎥
colorFrom: blue
colorTo: green
sdk: streamlit
python_version: 3.10
sdk_version: 1.10.0
suggested_hardware: "t4-medium"
suggested_storage: "medium"
app_file: app.py
app_port: 8501
base_path: /
fullWidth: true
header: default
short_description: Detect deepfakes and video manipulations using AI-based analysis.
models:
  - facebook/detr-resnet-50
tags:
  - deepfake-detection
  - video-analysis
  - AI
thumbnail: https://example.com/thumbnail.png
pinned: true
hf_oauth: false
hf_oauth_scopes: []
hf_oauth_expiration_minutes: 480
hf_oauth_authorized_org: []
disable_embedding: false
startup_duration_timeout: 30m
custom_headers: {}
preload_from_hub:
  - facebook/detr-resnet-50
Description
This Streamlit app allows users to upload a video, which is then analyzed for potential manipulation. It uses the following key features:

Frame Extraction: Captures frames from the video at defined intervals.
Object Detection: Uses the DETR model (a transformer-based object detection model) to detect objects in each frame.
Optical Flow Calculation: Computes motion patterns between consecutive frames to detect any abnormal motion (indicative of manipulation).
Manipulation Score: Based on the results of object detection and motion analysis, the app generates a manipulation score to indicate the likelihood of tampering.
Steps to Deploy
Clone the Repository:

git clone https://github.com/yourusername/video-manipulation-detection.git
cd video-manipulation-detection
Create a Virtual Environment: Create a virtual environment (optional but recommended):

python3 -m venv venv
Activate the virtual environment:

On Windows: venv\Scripts\activate
On Mac/Linux: source venv/bin/activate
Install Dependencies: Install the required dependencies:

pip install -r requirements.txt
Run the Streamlit App: To run the app locally:

streamlit run app.py
This will start the app on http://localhost:8501.

Deployment on Streamlit Space:

Push the repository to Hugging Face Spaces under your account.
Ensure the app.py file is located at the root of the repository.
The app will automatically run in the cloud on Streamlit.
Requirements
Python 3.10
Streamlit
OpenCV
PyTorch
Hugging Face Transformers
PIL (Python Imaging Library)
tqdm
Usage
Upload a video file (e.g., .mp4, .mov).
The app will process the video and display:
The manipulation score based on the likelihood of manipulation.
Frame-based analysis and optical flow insights.
View the final result on the app interface, indicating whether the video is likely manipulated.
License
This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments
DETR (facebook/detr-resnet-50) model from Hugging Face for object detection.
OpenCV for frame extraction and optical flow calculations.
Streamlit for creating the interactive web interface.

### Key Changes for Streamlit Space Deployment:

1. **`sdk: streamlit`**: Specifies that this is a Streamlit-based application.
2. **`app_file: app.py`**: Indicates the main file for the app (`app.py`).
3. **Deployment Steps**: Instructions on how to deploy the app on Streamlit, including how to clone the repository and install dependencies.
4. **Hugging Face Specific Configuration**: The configuration YAML ensures the app is correctly set up when deploying to **Hugging Face Spaces**.
   
Once this is set up, you'll be able to push your project to **Hugging Face Spaces**, where it will run as a **Streamlit** app.