Inspiration

We wanted to explore how computer vision could transform ordinary traffic camera footage into useful traffic insights without requiring expensive hardware. The goal was to create a smart system capable of detecting, tracking, and analysing vehicles automatically.

What We Built

TrafficVision detects and tracks vehicles in video streams, estimates their speed using camera calibration, recognises license plates with OCR, and generates annotated videos and CSV logs. The modular design allows future features such as traffic violation detection, dashboards, and database integration.

How We Built It

The project was developed in Python using OpenCV, Ultralytics YOLO, ByteTrack, EasyOCR, NumPy, and PyYAML. The pipeline processes each frame through detection, tracking, speed estimation, OCR, and output generation.

Challenges

The main challenges were achieving accurate speed estimation through camera calibration, maintaining stable tracking IDs during occlusions, and improving license plate recognition under varying lighting and camera angles.

What We Learned

This project strengthened our understanding of computer vision, multi-object tracking, OCR, pipeline design, and building scalable, modular software for real-world traffic analysis.

Built With

• bytetrack
• easyocr
• numpy
• opencv
• python
• ultralytics
• yolo