The bots follow the direction of motion of the four actuators of the body- the arms and the legs.

Computer vision

Used to track insect kinematics. Offline version of Kinefly that can run on Windows/Linux without ROS.

MotuS Machine Learning Repositories. Estimation rehabilitation treatment as metrics.

Scalable Health Care

Activity Detection using Pose Estimation and Object Detection

Example of pose estimation for bicep curl developed during Patch 2025

This project is a bowling game played using the Zed Camera Depth Sensor. The Zed Camera is a stereo camera that provides depth perception by capturing two synchronized images from two different viewpoints. It allows us to calculate the depth of objects in the scene and use that information to create interactive applications like games.

"Real-time multi-person 2D pose estimation using OpenPose with Caffe, implemented in Google Colab."

The case elaboration of making a demonstrator for the pose estimation demonstrator of third semester called "Innovate" at Zuyd University

Pytorch Code for paper "Fast Human Pose Estimation"

🤖 🃏 Teaching a CNN to estimate how many cards are in a pack.

Real-time pose estimation and action recognition

Computer Vision and Deep Learning

Pose2COCO Converter is a tool designed to transform pose annotations generated by OpenPose into the COCO format. This conversion is crucial for using pose data in COCO-based machine learning models and frameworks.

AI-driven solution can detect various workouts, monitor body orientation, count reps, and provide real-time feedback—all thanks to the power of computer vision and machine learning.

This project automates archery posture analysis. It processes video, detects poses, and analyzes movements, including detailed stance phase evaluation. It generates textual feedback, calculates scores, and creates feedback videos with overlaid corrections. Key features include saving individual feedback JSONs, CSV landmark export.

Models made for Edge Devices and NN Optimizations