PoseVision

Medical Tech track: Analyze squat posture in seconds and train with confidence.

Comment

Introduction

Proper squat form is crucial to avoiding injuries and maximizing results. This tool analyzes your squat video and highlights imbalances to show where you need to improve.

Inspiration

Squats are one of the most effective full-body exercises and are fundamental in strength training, yet they account for approximately 30% of weightlifting injuries, with the lower back and knees being particularly vulnerable. We created this project to help people recognize form issues early and train more safely.

What it does

This project analyzes squat videos to detect imbalances in knee angles, which can indicate potential issues with form. When the angles of the knees are different, the system highlights this in real-time, drawing attention to areas that may lead to injury. By providing clear visual feedback, it helps users identify posture concerns and improve their squatting technique.

How we built it

We built this project using Python and Flask for the back-end, leveraging OpenCV and MediaPipe for real-time video analysis. MediaPipe's pose estimation model helps us detect key joint positions, while OpenCV processes and highlights knee angle discrepancies. The front-end is powered by HTML and CSS, providing a clean and user-friendly interface for video upload and feedback display.

Challenges we ran into

The biggest challenge was developing an accurate algorithm to detect uneven squat posture, particularly in assessing knee angle discrepancies. We had to fine-tune the pose estimation model from the deep learning library MediaPipe to reliably track joint positions in real-time. After extensive testing and optimization, we successfully created an algorithm that can detect these posture imbalances, providing users with actionable visual feedback.

Accomplishments that we're proud of

We are proud to have used deep learning and computer vision techniques to accurately analyze squat posture in real-time. By leveraging advanced algorithms, we’re able to detect even the smallest imbalances in knee angles, helping users spot potential issues with their form. This innovative use of technology allows us to offer personalized suggestions for improvement, ultimately helping reduce the risk of injury.

What we learned

Throughout the project, we learned how to use OpenCV, a powerful computer vision library, to process and manipulate video frames in real-time. We also gained experience with MediaPipe, a framework that offers state-of-the-art pose estimation models for tracking body joints and movements. By combining these tools, we were able to accurately analyze and track squat posture to detect potential form issues.

What's next for PoseVision

Next, we plan to expand the project to analyze other key exercises, starting with deadlifts and bench presses, to provide comprehensive feedback on various lifting techniques. By leveraging the same computer vision and pose estimation technologies, we aim to detect common posture issues in these exercises, such as improper back alignment during deadlifts or hand positioning during bench presses. This will help make the tool more versatile and valuable for users looking to improve their overall workout safety and performance.

Built With

Share this project:

Updates