Reproduce “Learning Association via Track-Detection Matching for 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
Momir Adžemović
TDLP learns to predict links between existing tracks and new detections in a modular video tracker, replacing hand-designed association rules. It encodes track histories and detections, scores each pair, and matches likely links. On SportsMOT’s test set, it achieves 81.9 HOTA versus CAMELTrack’s 80.4, but scores below leading methods on MOT17; its measured throughput is also lower than CAMELTrack’s.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.