Browse Teach Curate

Cue Fusion Methods for Data Association in Multi-Object Tracking

How the per-pair association costs in tracking-by-detection multi-object tracking can be combined from several cues - strong ones like motion and appearance and weak ones like height-IoU and tracklet confidence - and how the operator used to fuse them changes matching quality. Students learn four fusion operators (minimum, IoU-weighted sum, Kalman-filter gating, and Hadamard product) and how to choose among them.

KITT Text

FusionSORT: Fusion Methods for Online Multi-object Visual Tracking Nathanael L. Baisa

An investigative study comparing four ways to fuse the association costs used in tracking-by-detection multi-object visual tracking, where object detections in each video frame are matched to existin…