GeezerVison

YouTube Video Frame Extraction & Analysis

This project extracts frames from YouTube videos and performs analysis on the extracted frames. Currently, the system supports color analysis, face detection, and scene change detection.

• Overview
• Features
• Setup
• Usage
• Frame Analyses
• Folder Structure
• Results
• Future Improvements
• Contributions

The project automates the extraction of frames from a provided YouTube video URL. Once the frames are extracted, analyses such as color analysis, face detection, and scene change detection are performed on them. This project aims to categorize video content and provide insight into the structure of video frames.

• Frame Extraction: Extracts frames from YouTube videos (ignores ads).
• Color Analysis: Analyzes dominant colors in each frame.
• Face Detection: Detects faces in each frame using OpenCV’s Haar cascades.
• Scene Change Detection: Detects significant scene transitions in the video.

git clone https://github.com/your-username/Youtube-Video-Frame-Extraction-Analysis.git
cd Youtube-Video-Frame-Extraction-Analysis

python3 -m venv yt_env
source yt_env/bin/activate   # For Linux/Mac
yt_env\Scripts\activate      # For Windows

pip install -r requirements.txt

• Mac: brew install tesseract
• Ubuntu: sudo apt install tesseract-ocr
• Windows: Download the Tesseract installer

Run the frame extraction script by providing a YouTube video link:

python youtube_frame_extractor.py "https://www.youtube.com/watch?v=VIDEO_ID"

Frames will be saved in the frames/ directory.

You can run different analysis scripts on the extracted frames:

• Color Analysis:

  python color_analysis.py

• Face Detection:

  python face_detection.py

• Scene Change Detection:

  python scene_change_detection.py

Detect dominant colors in each frame and store the results in results/color_analysis/.

Detect human faces in the frames and save annotated images with bounding boxes in results/face_detection/.

Identifies and logs frames where a scene change occurs, and the results are stored in results/scene_change/.

Youtube-Video-Frame-Extraction-Analysis/
├── frames/                   # Extracted frames from the video
├── results/                  # Analysis results
│   ├── color_analysis/
│   ├── face_detection/
│   └── scene_change/
├── youtube_frame_extractor.py # Main script to extract frames
├── color_analysis.py          # Script for color analysis
├── face_detection.py          # Script for face detection
├── scene_change_detection.py  # Script for scene change detection
├── requirements.txt           # Python dependencies
└── README.md                  # This file

Each analysis stores results in a separate folder inside the results/ directory. This keeps the outputs of different types of analysis organized.

• Object Detection: Add support for detecting objects using models like YOLOv3 or YOLOv4.
• Text Detection (OCR): Implement text extraction from frames using Tesseract OCR.
• Motion Detection: Add the ability to detect motion across consecutive frames.
• Video Summarization: Integrate summarization techniques to highlight key moments in the video.

Feel free to submit pull requests for new features, bug fixes, or improvements. Any contributions are welcome!