Problem:
A leading out-of-home (OOH) advertising agency manages thousands of billboards across Malaysia, and needed their existing process of quarterly tracking and revenue reporting to be automated to:
- Accurately detect brands from billboard images
- Eliminate manual mapping and reduce human error
- Generate timely and accurate revenue forecasts
- Manual tracking was slow, inconsistent, and lacked real-time insights.
Approach:
- Billboard Detection and Mapping: Applied computer vision models to automatically detect and isolate billboard content from field photos. Extracted geolocation and orientation data from image metadata for precise mapping to the correct billboard asset.
- Brand Identification: Applied OCR and NLP techniques to extract brand names from the billboards and developed custom matching logic for accurate brand detection.
- Automated Revenue Calculation: Mapped detected brands to predefined billing cycles, enabling accurate revenue estimation.
Impact:
- Real-time visual dashboards and faster, automated revenue reports that drive ad presence
- Accurate brand presence detection across billboards and reduced human intervention by 80%
- Location-based performance insights for campaign planning and audit purposes
- Brand matching accuracy went up to 90%
