Multi-Object Tracking and Re-Identification
Following people, vehicles or animals across video frames and cameras and recognising them again.
Tier Grades
Tier A: Frontier AI labs
feasible
Academic and industry labs publish trackers such as ByteTrack.
Tier B: Indonesian AI labs
feasible
A tracker that runs at 30 FPS on one V100 is within an Indonesian lab reach.
Tier C: Institution units such as ITB
feasible
Benchmarking on MOT17 and MOT20 fits a university unit.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
feasible
Tracking turns detections into counts and flows for traffic, retail and plantation pilots.
Where this comes from
Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.
Anchors
- ComputeByteTrack runs at 30 FPS on a single V100 GPU on MOT17.
- DataMOT17, MOT20, HiEve and BDD100K tracking benchmarks.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 9 authors on the ByteTrack paper (arXiv 2110.06864).
- Time horizonYears.
- Shelf-life signalsTrackers fold into detection models, Privacy rules limit cross-camera re-identification of people
- Pipeline roleTier A and academia publish trackers; Tier C tests them on local footage; Tier D uses them for counting and flow analysis.
Where this comes from
Team size is a proxy (author counts of exemplar papers), not a headcount.
Who and What
Exemplars
Observed actors
Huazhong University of Science and Technology
University of Surrey
Queen Mary University of London
Carnegie Mellon University
Shanghai AI Lab
NVIDIALinked models
OC-SORTvisionexemplarByteTrackvisionexemplarOSNetvisionexemplarDeepSORTvisionexemplar
EdgeTAMvisionproduced by
MegaDescriptorvisionproduced by
Track-Anythingvisionproduced by
OSTrackvisionproduced by
MixFormervisionproduced byStrongSORTvisionproduced by
STARKvisionproduced by
QDTrackvisionproduced byFairMOTvisionproduced by
PrDiMPvisionproduced by
MaskTrack R-CNNvisionproduced by
Tracktorvisionproduced by
SiamRPN++visionproduced by
AlphaPosevisionproduced by
SPGANimage generationproduced bySORTvisionproduced by
Intel person attributes and re-identificationvisionproduced byReIdentificationNetvisionproduced byVehicle ReID 0001visionproduced byLinks
R&D TopicMulti-Object Tracking and Re-Identification
ModelsOSNetUniversity of Surrey, visionDeepSORTvisionByteTrackHuazhong University of Science and Technology, visionOC-SORTCarnegie Mellon University, vision+19 more
OSNetUniversity of Surrey, visionDeepSORTvisionByteTrackHuazhong University of Science and Technology, visionOC-SORTCarnegie Mellon University, vision+19 moreUse Casesnone recorded
Who Runs Itnone recorded
Models 23
MaskTrack R-CNNByteDance, University of Illinois Urbana-Champaign, Adobe Inc., visionproduced byEdgeTAMMeta, visionproduced byOC-SORTCarnegie Mellon University, Shanghai AI Lab, NVIDIA, visionexemplarOSNetUniversity of Surrey, Queen Mary University of London, visionexemplarSPGANUniversity of Chinese Academy of Sciences, University of Technology Sydney, image generationproduced byAlphaPoseShanghai Jiao Tong University, visionproduced byFairMOTHuazhong University of Science and Technology, Microsoft Research, visionproduced byIntel person attributes and re-identificationIntel, visionproduced byMegaDescriptorCzech Technical University in Prague, visionproduced byOSTrackInstitute of Computing Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, visionproduced byQDTrackZhejiang University, ETH Zurich, visionproduced bySiamRPN++SenseTime, Institute of Automation, Chinese Academy of Sciences, visionproduced by+11
STARKDalian University of Technology, Microsoft Research, visionproduced byTrack-AnythingSouthern University of Science and Technology, visionproduced byByteTrackHuazhong University of Science and Technology, visionexemplarMixFormerNanjing University, visionproduced byPrDiMPETH Zurich, visionproduced byReIdentificationNetNVIDIA, visionproduced byTracktorTechnical University of Munich, visionproduced byDeepSORTvisionexemplarSORTvisionproduced byStrongSORTvisionproduced byVehicle ReID 0001visionproduced byPapers
- Simple Online and Realtime TrackingAlex Bewley et al., arXiv preprint, 2016 founding
- ByteTrack: Multi-Object Tracking by Associating Every Detection BoxYifu Zhang et al., arXiv preprint, 2021 state of the art
- FairMOT: On the Fairness of Detection and Re-Identification in Multiple Object TrackingYifu Zhang et al., International Journal of Computer Vision, 129(11), 2021, 3069-3087, 2020 state of the art
- Simple Online and Realtime Tracking with a Deep Association MetricNicolai Wojke et al., arXiv preprint, 2017 state of the art
- Multiple Object Tracking: A Literature ReviewWenhan Luo et al., arXiv preprint, 2014 survey
Where this comes from
The papers to read first: where the direction began, its current best, and a survey when one exists.
Sources
- arxiv.org/abs/2110.06864 https://arxiv.org/abs/2110.06864
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Multi-Object Tracking and Re-Identification", https://aurion.semai.id/topics/multi-object-tracking-and-reidentification/, records of October 2026.