Reproduce “Towards Discriminative Representation: Multi-view Trajectory Contrastive Learning for Online Multi-object Tracking” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
En Yu, Zhuoling Li, Shoudong Han
MTrack tracks objects across video frames by combining adaptively selected appearance points with a compact memory of trajectories. Its training contrasts appearance vectors with trajectory centers, and it downweights inconsistent features at inference. On the MOT17 private-detection test set, it scores 73.5% IDF1 and 72.1% MOTA, versus QDTrack’s 68.7% and 66.3%; trajectory-filling post-processing means this comparison does not isolate the learned representation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.