AI Paper Atlas

Research map / Detection and segmentation / Tracking

ByteTrack: Multi-Object Tracking by Associating Every Detection Box

Yifu Zhang, Peize Sun, Yi Jiang, Dongdong Yu, Fucheng Weng, Zehuan Yuan, Ping Luo, Wenyu Liu, Xinggang Wang

ECCV 2022 · Detection and segmentation › Tracking · approach: Association cues for online MOT · 3,110 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add second association of low-score boxes (BYTE) to SORTSORT · MOT17 val (half)MOTA: 74.6 -> 76.6Improved
add low-score association to SORTSORT · MOT17 val (half)IDs: 291 -> 159Improved
use IoU instead of Re-ID as similarity in second association (ablation)BYTE · MOT17 val (half)MOTA: 75.8 -> 76.6Improved
use Re-ID instead of IoU in first association on BDD100K (ablation)BYTE · BDD100K valmMOTA: 39.4 -> 45.5Improved
use Re-ID in first association instead of IoU (ablation)BYTE · MOT17 val (half)MOTA / IDF1: MOTA 76.6 -> 76.3, IDF1 79.3 -> 80.5Mixed
apply BYTE (Kalman) to CenterTrackCenterTrack · MOT17 val (half)IDF1: 64.2 -> 74.0Improved
apply BYTE to Chained-TrackerCTracker · MOT17 val (half)IDF1: 60.9 -> 66.7Improved
insert BYTE into original association of MOTRMOTR · MOT17 val (half)MOTA / IDF1: MOTA 64.7 -> 64.3, IDF1 67.2 -> 69.3Mixed
replace TraDes association with BYTE (Kalman only)TraDes · MOT17 val (half)MOTA / IDF1: MOTA 68.2 -> 67.9, IDF1 71.7 -> 72.0No clear effect
insert BYTE into FairMOT associationFairMOT · MOT17 val (half)MOTA: 69.1 -> 70.4Improved
replace DeepSORT with BYTE on YOLOX-NanoYOLOX-Nano detector · MOT17 val (half)MOTA: 61.4 -> 64.4Improved
add tracklet interpolation (sigma 20) (ablation)ByteTrack · MOT17 val (half)MOTA: 76.6 -> 78.3Improved

Later papers that compared with this method

38 later accepted papers report ByteTrack as a baseline in their results tables; 36 of them report a better number on most of the results they share with it.

30 more not listed.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy