Highway traffic counting for outdoor advertising analytics

The challenge

Advertisers wanted evidence of how many vehicles actually passed each billboard, but there was no reliable, continuous count.

What we built

• Vehicle frames annotated by class (car, bus, truck, two-wheeler) to fine-tune a pretrained YOLOv8 model on local traffic.
• Frame-skip optimization processes every fourth frame to handle high-FPS video efficiently.
• Centroid tracking keeps unique vehicle IDs across frames so each vehicle is counted once.
• Class-wise counts exposed through an API to feed hourly and daily traffic-density reports.

Results

• Evidence-based impressions data for media planning
• Benchmarking of billboard visibility and reach
• Automated reporting with no manual counting