posture-insights-vision

Real-time exercise form correction using YOLOv8 pose estimation — your personal AI fitness assistant.

Overview

Posture Insights Vision is a computer vision-based system designed to provide real-time feedback on exercise posture and form. Leveraging YOLOv8's pose estimation, this project helps users self-correct during workouts by identifying improper alignment and joint angles, all through a simple web interface powered by Flask and OpenCV.

Whether you're doing push-ups, planks, or curls, this tool acts as your virtual personal trainer to ensure you're doing it right.

How It Works

1. Web Interface Launch: Users open a Flask-powered web interface and select an exercise.

2. Webcam Activation: The app activates the webcam to begin capturing live video.

3. Pose Estimation: Each frame is passed through YOLOv8's pose estimation model, detecting 17 key body joints in real-time.

4. Heuristic Analysis: Based on selected exercise, relevant joints are analyzed for:

   • Joint angles

   • Body alignment

   • Motion flow

5. Real-Time Feedback: If poor posture is detected, alerts are displayed on the UI instantly.

Sample Output

Features

• ✅ Real-time posture detection

• 🎯 Accurate joint tracking with YOLOv8

• 🧩 Modular exercise-specific heuristics

• 💻 Lightweight Flask app with webcam streaming

• 📈 Designed to work on local machines with webcam access

Supported Exercises & Heuristics

Each heuristic is designed using joint angle thresholds and motion trajectory tracking based on keypoints detected per frame:

1. Plank – Ensures shoulder, hip, and ankle are aligned in a straight line.

2. Push-up – Tracks elbow flexion and extension, back curvature, and wrist alignment.

3. Bicep Curl – Measures elbow joint angle to ensure full range of motion and controlled return.

Getting Started

1. Clone the Repository:

   git clone https://github.com/RohitKrish46/posture-insights-vision.git
   cd posture-insights-vision
   

2. Create a Virtual Environment:

   python -m venv env
   

3. Activate the Virtual Environment:

   • On Windows

     env\Scripts\activate
     

   • On macOS and Linux

     source env/bin/activate
     

4. Install Dependencies:

   pip install -r requirements.txt
   

5. Download the YOLOV8 Model:

   • Download the desired YOLOV8 model and store it under the 'models' folder.

   • You can download ultralytics models using:

     from ultralytics import YOLO
     model = YOLO('yolov8m-pose.pt')
     

6. Launch the Flask App:

   flask run
   

7. Open in Browser:

   Navigate to http://127.0.0.1:5000 to start using the app.

Playing with YOLOv8's Pose Estimation

YOLOv8's pose model is ideal for this use case due to:

• Single-pass detection of all keypoints

• High FPS for real-time inference

• Strong accuracy for joint localization

The model outputs coordinates for 17 joints, which are further processed using geometric and motion-based heuristics.

Project Structure

posture-insights-vision/
│
├── app.py                   # Flask app entry point
├── models/                  # YOLOv8 model files
├── static/                  # CSS and JS files
├── templates/               # HTML templates
├── utils/
│   └── heuristics.py        # Heuristic functions for posture detection
├── requirements.txt
└── README.md

Future Enhancements

1. Add more exercises (e.g., squats, lunges)

2. Add feedback history and performance metrics

3. Integrate with fitness APIs for tracking

4. Port to mobile or desktop apps