YOLO brand detection for sponsor analytics

For sponsor analytics at DosBro, I developed a YOLOv4-based brand detection framework that processed 300K+ social media images per month and replaced a large chunk of manual tagging work.

The system delivered 89% mAP and cut manual labeling effort by 40%, giving marketing teams consistent brand exposure data across Instagram and similar platforms without waiting on human review for every image.

Focus

  • YOLOv4 object detection at social-media scale
  • Sponsor and brand analytics automation
  • High-volume image processing pipelines