Instant billboard replacement in any video—powered by advanced computer vision. No camera tracking. No legacy "virtual ad" tech. Just pure AI.
The POC you see now is the result of testing, breaking, and re-building over 400+ code versions and approaches.
Every broadcast, every stadium, and every lighting condition brought its own challenge.
The real secret? Relentless engineering and field validation. There is no magic script – just the best of modern AI, computer vision, and a lot of patience.
We trained our own deep learning models (Mask R-CNN variants) to detect advertising boards with pixel-level accuracy.
Why not use ready models? Because every stadium and league is different. Our network was tailored for real, messy sports video.
The “magic” isn’t just detection. We use fast, GPU-based tracking (RAFT & custom logic) to follow each board over time, enabling smooth transitions and perspective correction even with fast camera moves and scene cuts.
To prevent errors, we run parallel models for humans, balls, and other moving objects (COCO / pretrained Mask R-CNN).
Billboards are only replaced when we’re 100% sure it’s safe — this is not a toy, it’s production-ready AI.
The entire pipeline — from detection to mask blending and video rendering — runs on CUDA and GPU acceleration (PyTorch, TensorRT).
No CPU bottlenecks, no off-the-shelf “scripts.”
Each step — from basic detection, to robust tracking, to multi-model masking — was validated on thousands of real frames, not synthetic test sets.
Many attempts failed. The result is what survived real-world chaos.
Most “sample scripts” or public code crashed or lagged at broadcast scale.
Our code is 95% custom, built for real performance and real clients.
The real world is messy. We tuned our pipeline to work across formats, resolutions, and camera angles.
That’s why you can try your own video in the demo, not just canned samples.
What you see is just the start. With more data, feedback, and resources, we can make it faster, smarter, and globally scalable.
We did the hard work to get AI billboard replacement working in the wild.
Now, with the right partners, we can take it to broadcast scale, integrate with leading video platforms, and automate the future of sports advertising.