search

MasterBin-IIAU / Unicorn

[ECCV'22 Oral] Towards Grand Unification of Object Tracking
MIT License
949 stars 87 forks source link

MOT模式下embedding的loss与qdtrack的loss是一样的么？ #35

Closed long2double closed 2 years ago

MasterBin-IIAU commented 2 years ago

Hi, for simplicity when computing the embedding loss, we only use sparse GTs instead of quasi-dense proposals like QDTrack.