Computer Vision Toolkit

Overview

A suite of 6+ real-time computer vision modules built with OpenCV — covering edge detection, object tracking, and video stabilization, all running on live webcam feeds.

The Problem

Computer vision concepts are often taught in isolation. The goal was to build a cohesive, modular toolkit where each technique builds on a shared foundation and can be tested live on a webcam feed.

My Role

Sole author. Designed the modular architecture and implemented each vision module independently with shared utility code.

Key Features

• —Edge detection with configurable Canny thresholds
• —Real-time object tracking with bounding box rendering
• —Video stabilization for handheld/shaky footage
• —Modular architecture — each module is independently runnable
• —All processing runs at <100ms latency on webcam feeds

Challenges

Maintaining acceptable frame rates across all modules simultaneously. Profiled each pipeline and optimized the NumPy operations that were creating per-frame bottlenecks.

What I Learned

Profiling matters more than algorithmic cleverness for real-time systems. The bottleneck is almost never where you expect it to be.