Computer Vision

Computer Vision

Overview

I began my computer vision journey at Calivision through practical learning, where I developed software solutions for monitoring and tracking gym workouts using computer vision technology. My work involved leveraging multi-camera systems to provide seamless, real-time tracking of individuals, enhancing the accuracy and effectiveness of movement analysis in dynamic gym environments.

Key Skills & Technologies

• Programming: Python
• Frameworks/Libraries: OpenCV, YOLO
• Tools: VS Code
• Concepts: Multi-Camera Calibration, Object Detection, Real-Time Tracking, Ai integration, Occlusion Handling & Enhanced Tracking

Notable Projects

1. Multi-Camera Workout Tracking System
   • Accurate Athlete Monitoring: Developed an advanced multi-camera tracking system, enabling precise monitoring of individuals across multiple camera streams.
   • Multi-Camera Calibration: Implemented a calibration process to align different camera views, ensuring consistency in the bounding boxes generated by YOLO (You Only Look Once) across multiple camera angles.
   • Real-Time Tracking Integration: Integrated real-time tracking to enable continuous monitoring and accurate representation of individuals’ movements during workouts, across different perspectives.
2. Occlusion Handling & Enhanced Tracking (In Progress)
   • Research & Development: Currently researching and developing advanced techniques to handle occlusions (object blocking) and reduce false positives in multi-camera tracking systems.
   • Accuracy & Reliability Focus: Working on improving the system’s accuracy and reliability by addressing challenges related to object interference and misidentification.

Achievements & Impact

• Improved Tracking Accuracy: Successfully improved tracking accuracy by implementing effective multi-camera calibration, aligning bounding boxes across different camera streams.
• Enhanced Real-Time Monitoring: Elevated real-time monitoring capabilities for gyms, providing seamless tracking of athletes from various angles, enhancing user experience and operational efficiency.

Future Goals & Learning

I’m excited to explore occlusion handling techniques and refine tracking models to ensure more reliable and seamless monitoring across diverse camera views. Additionally, I plan to integrate deep learning-based pose estimation to further enhance movement analysis in gym environments.