Pinned
labelling real-world videos for point tracking is cumbersome. therefore, we use synthetic training data, but this creates a gap with real-world test videos.
what if we could use existing trackers to label real-world videos? but this requires knowing which tracker to trust when.
01:36
Happy to share our work: Real-World Point Tracking with Verifier-Guided Pseudo-Labeling. #CVPR2026
We improve the pseudo-label training pipeline for real-world videos using a verifier that selects the most reliable predictions across multiple trackers.
🔗kuis-ai.github.io/track_on_r
