HomeProjects
← Back to Projects

YOLO Liga

Sponsor Detection in Football Broadcasts

Developed a computer vision system to automatically detect sponsor advertisements in Portuguese football broadcasts, giving companies a data-driven way to measure their broadcast ROI.

Benchmarked YOLOv8 (s, m) and YOLO11 (m, l) on a dataset of 3,000+ images. YOLO11l, trained on a cloud NVIDIA RTX 6000 Ada, achieved a mAP50-95 of 0.73 — sufficient for production deployment in real-time stream analysis.

Dataset

Source:
Portuguese Liga broadcasts
Size:
3,000+ annotated images
Task:
Object detection of sponsor advertisements (billboards, LED boards, overlays)

Models Benchmarked

  • YOLOv8s (small)
  • YOLOv8m (medium)
  • YOLO11m (medium)
  • YOLO11l (large)

Best Model: YOLO11l

Architecture:
YOLO11 Large
mAP50-95:
0.73
Training Hardware:
NVIDIA RTX 6000 Ada (Cloud)
Deployment:
Real-time stream analysis ready

Training Results

YOLO11l training results showing loss curves and mAP metrics

Use Case

This system enables sponsors and broadcasters to automatically measure advertisement visibility during live football matches. By detecting when and how long sponsor logos appear on screen, companies can quantify their broadcast ROI with precise, frame-by-frame data instead of relying on manual estimates.

Sample Output

Full frame-by-frame sponsor detection report for the second half of Sporting CP vs. AFS (Liga Portugal Betclic, Matchday 14, 14/12/2025).

↓ Download Analysis (XLSX, 7.9 MB)