Overlander26 -Human Pose Detection

Demo - "Hello World"

• Choose YouTube video , view and download (Streamlit UI)
• Get vid with Human movement (face not always visible - above the virtual fence line). As and when face visible (Left-EYE Visible) , grab that Image Frame for further processing.

📸 Screenshots & Demo Videos

• Original Video - https://www.youtube.com/shorts/TWNxXXeLmM0

Type	Filename	Preview	Description
🖼️ Image	Face-Image-1		(Left-EYE)
🖼️ Image	Face-Image-2		(Left-EYE Occluded)
🖼️ Image	Face-Image-3		(Left-EYE Occluded)

---

Type	Filename	Preview	Description
🖼️ Image	Screenshot from 2026-02-14 21-41-53.png		Initial screenshot of application
🎥 Video	gym_1.mp4	▶️ View Video	Gym pose detection demo video (1006K)
🖼️ Image	frame_0002.jpg		Extracted frame from video (70K)
🖼️ Image	frame_0002.jpg_frame_pose_46__0__.png		Pose landmarks detected (timestamp 46s)
🖼️ Image	frame_0002.jpg_frame_pose_52__0__.png		Pose landmarks detected (timestamp 52s)

---

🎯 Project Overview

This project implements human pose detection using MediaPipe Tasks API 0.10+ with the following features:

• ✅ Real-time pose landmark detection
• ✅ Support for video files (MP4, AVI, MOV, MKV)
• ✅ Support for live camera feeds (RTSP, USB webcam)
• ✅ MediaPipe Tasks API 0.10+ (modern, optimized)
• ✅ 25-30 FPS performance (3-5x faster than deprecated API)

---

📊 Demo Results

Frame Processing Example

Original Frame

Pose Detection Result

---

🚀 Quick Start

# Clone repository
git clone https://github.com/YourUsername/overlander26.git
cd overlander26

# Create virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Download MediaPipe model
mkdir -p data_dir/pose_models
wget -O data_dir/pose_models/pose_landmarker.task \
  https://storage.googleapis.com/mediapipe-models/pose_landmarker/pose_landmarker_heavy/float16/1/pose_landmarker_heavy.task

# Run pose detection
python main.py

---

📁 Directory Structure

overlander26/
├── screen_shots/          # Demo images and videos
├── data_dir/
│   ├── pose_models/       # MediaPipe model files
│   └── pose_detected/     # Output frames
├── src/
│   └── analysis/
│       └── media_pipe.py  # Pose detection implementation
└── README.md

---

🎥 Video Demo

Gym Pose Detection Demo:

Click the image above to view the full demo video

---

Tech Stack

• MediaPipe Tasks API 0.10+ - Pose landmark detection
• OpenCV - Video processing and frame capture
• NumPy - Array operations

---

Features

• MediaPipe Tasks API 0.10+ (modern, optimized)
• Video file processing (MP4, AVI, MOV, MKV)
• Live camera feed support (RTSP, USB, IP camera)
• Automatic codec detection and error handling
• Pose landmark visualization
• Frame-by-frame analysis
• High performance (25-30 FPS)

---