PIPs: Particle Video Revisited: Tracking Through Occlusions Using Point Trajectories

• Model Authors: Adam W Harley and Zhaoyuan Fang and Katerina Fragkiadaki
• Paper: Particle Video Revisited: Tracking Through Occlusions Using Point Trajectories (ECCV 2022 - https://arxiv.org/abs/2204.04153
• Code Repo: https://github.com/aharley/pips
• Project Homepage: https://particle-video-revisited.github.io

From the paper abstract:

[...] we revisit Sand and Teller's "particle video" approach, and study pixel tracking as a long-range motion estimation problem, where every pixel is described with a trajectory that locates it in multiple future frames. We re-build this classic approach using components that drive the current state-of-the-art in flow and object tracking, such as dense cost maps, iterative optimization, and learned appearance updates. We train our models using long-range amodal point trajectories mined from existing optical flow data that we synthetically augment with multi-frame occlusions.

Citation

@inproceedings{harley2022particle,
  title={Particle Video Revisited: Tracking Through Occlusions Using Point Trajectories},
  author={Adam W Harley and Zhaoyuan Fang and Katerina Fragkiadaki},
  booktitle={ECCV},
  year={2022}
}