Dynamic motion detection with human pose estimation

Project Overview:

The project revolves around leveraging human keypoint datasets to train a model capable of identifying joint positions (such as shoulders, elbows, and knees) and connecting them to form a virtual skeleton representing the body’s posture.

Key Features and Functionalities:

Angle Calculation: Compute the angles for each joint in the human body.

Single-Pose Estimation: Analyze and extract details from a single pose.

Multi-Pose Estimation: Accurately analyze multiple poses in an image.

Video Pose Estimation: Process video files seamlessly to gain insights into body motion.

Live Stream Pose Estimation: Real-time pose detection via webcam or external camera.

Fall Detection: Identify falls in real-time, with potential applications in elderly care, workplace safety, and healthcare monitoring.

User-Friendly Interfaces:

We developed multiple interactive interfaces to ensure smooth user experiences:

Web Application: Upload images or use live stream input to visualize results with the virtual skeleton overlay.

Windows Application: Built with custom Tkinter, this app offers various functions accessible through dedicated buttons:

Single-Image Pose Estimation: Analyze one person’s pose and calculate joint angles.

Multi-Image Pose Estimation: Analyze multiple individuals in an image with precise joint angle measurements.

Video Pose Estimation: Smooth processing for video-based motion analysis.

Webcam Pose Estimation: Real-time skeleton overlay and joint analysis using a camera.

Future Development:

We aim to expand this project, as it’s not just about technology—it’s about making meaningful contributions. From fitness enthusiasts to healthcare providers, the potential applications are endless.