COMPUTER VISION PROJECTS
Real-Time AI Personal Trainer
Spearheaded an advanced computer vision initiative for fitness applications, leveraging Google's MediaPipe for precise pose detection and OpenCV for an AI-driven fitness assistant. Developed a sophisticated system capable of recognizing diverse exercises, applying intricate pose detection algorithms seamlessly integrated with OpenCV. Demonstrated proficiency in pose identification, real-time feedback mechanisms, and advanced machine learning techniques, showcasing technical acumen at the intersection of computer vision and fitness innovation.
Semantic Segmentation of Street Features for Autonomous Vehicles
Engineered a state-of-the-art U-Net architecture featuring height-based attention using PyTorch for highly precise Semantic Segmentation of urban street elements on the challenging Cityscapes dataset. This project significantly contributes to the advancement of autonomous driving technology, enabling accurate identification of traffic lights and enhancing pedestrian detection. Attained an 86% pixel accuracy and a Mean Intersection over Union (IoU) score of 73%, showcasing a meticulous approach to advancing computer vision solutions for complex urban environments.
Gesture-based Volume Control
Implemented a computer vision project enabling gesture-based control of system volume. Leveraging advanced computer vision algorithms and frameworks, including OpenCV and MediaPipe, the system accurately interprets user gestures in real time. This innovative solution integrates seamlessly with the system's audio controls, providing an intuitive and hands-free experience, highlighting proficiency in computer vision, real-time processing, and system integration.