New AI research from Meta – CoTracker3 Simpler and Better Point Tracking by Pseudo-Labelling Real Videos.
More details ➡️ go.fb.me/xiyc63
Demo on @huggingface ➡️ go.fb.me/yzuqd0
Building on our previous work on CoTracker, this new model demonstrates impressive tracking results where points can be tracked for a long time even when they're occluded or leave the field of view. CoTracker3 achieves state-of-the-art, outperforming all recent point tracking approaches on standard benchmarks — often by a substantial margin.
We've released the research paper, code and a demo on Hugging Face — along with models available under an A-NC license to support further research in this space.
43
307
1,804
219,123
Meta's CoTracker3 sounds like a major advancement in point tracking! The ability to track points accurately, even when they're occluded or leave the field of view, is a huge leap forward. It’s exciting to see it outperform recent point tracking methods by a substantial margin. With the paper, code, and demo released on Hugging Face, this opens up big opportunities for further research and innovation. Can't wait to see how this impacts computer vision and related fields!
Check out the details here: go.fb.me/xiyc63 and demo it here: go.fb.me/yzuqd0
Oct 17, 2024 · 10:02 AM UTC
1
487

