Continental freight monitoring from public highway camera feeds

The problem

Freight movement between industrial areas and warehouses is visible on thousands of public highway cameras, but only as raw video. Turning that into a usable movement signal meant solving acquisition, cost and accuracy at the same time.

What we built

1. Collected continuous video feeds from government transportation sites using Selenium for acquisition and BeautifulSoup for parsing.
2. Converted feeds to sampled image frames, cutting compute cost sharply while preserving the movement signal.
3. Developed and deployed a YOLOv8 model for real-time detection and counting of trucks and containers.
4. Automated the processing and analysis pipeline with AWS Lambda, removing manual intervention and letting the system scale across states.
5. Analysed traffic flow patterns and integrated the resulting data pipeline into existing transport management systems for route optimisation.