Video Landmark Processing

This project provides tools for processing videos to detect and work with facial landmarks. Example vidoes contain one speaker saying "Van" and "Fan". The output video contains the landmarks of the speaker saying "Fan" side to side with the landmarks of the speaker saying "Van".

Features

• Facial landmark detection from video files
• Landmark-based face cropping
• Support for video processing and frame extraction

Setup

1. Install the required dependencies:

Usage

Generate Landmarks

Use generate_landmarks.py to detect and extract facial landmarks from a video:

Crop Video Using Landmarks

Use landmark_crop.py to crop video frames based on detected facial landmarks:

Project Structure

• generate_landmarks.py: Script for detecting and saving facial landmarks from videos
• landmark_crop.py: Tool for cropping videos based on landmark positions

Requirements

• Python 3.7+
• OpenCV
• MediaPipe (for facial landmark detection)
• NumPy

License

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