We are creative, ambitious and ready for challenges! Hire Us
PERSONALISED MEMORIES, MARATHON COVERAGE WITH AI DETECTION
Client’s Need
Autonomously detect and track registered athletes in marathon race video feeds captured at various locations. Aim is to generate and monetize personalized short videos and photograph reels for each participant without extensive / specialty hardware.
Goals of the project
- High accuracy rate athlete detection and tracking across different video feeds.
- Extract personalized video clips from each feed and stitch them together.
- Processing on standard Intel i7 machine without GPU.
- Eliminate manual feed capture and editing workforce.
Our Approach
Detect and correlate face / bib# combo for accurate athlete tracking.
Identify entry and exit frames for each participant across feeds.
Optimize code and model for speed and accuracy, with no need for GPU.
Outcome
97% accuracy in detecting and tracking athletes, even when the bib numbers were occluded using correlation to generate personalized continuous videos.
Real time processing of video feed on standard Intel i7 computers.
300% increase in client productivity with 30 minutes processing and offers for personalized video memorabilia to athletes.
Increased clientele, with more competitive offers to athletes.
Published:
September 28, 2024
Category:
Computer Vision
Client:
Oceanthemes
