前言
这段时间在整理毕设,所以这里结合了SpyderXu分享的内容把多目标跟踪相关的文献资源共享一下,由于文章很多,所以我这里只整理3年以内的,对于年限久远的,这里只取提供了代码的和比较经典的。并且尽可能注释了相关算法在MOT数据集上的名称。各自算法的性能比较可以看论文以及MOT官网。
在线跟踪(Online)
Name | Source | Publication | Notes |
---|---|---|---|
Adopting Tubes to Track Multi-Object in a One-Step Training Model | [pdf] [code] | CVPR2020 | TubeTK |
Joint Detection and Multi-Object Tracking with Graph Neural Networks | [pdf] | arxiv(2020) | JDMOT_GNN |
Graph Networks for Multiple Object Tracking | [pdf] [code] | WACV2020 | GNMOT |
Deep association: End-to-end graph-based learning for multiple object tracking with conv-graph neural network | [pdf] | ICMR2019 | DAN |
SQE: a Self Quality Evaluation Metric for Parameters Optimization in Multi-Object Tracking | [pdf] | arxiv(2020) | SQE |
Autoregressive Trajectory Inpainting and Scoring for Tracking | [pdf] | CVPR2020 | ArTIST |
Multiple Object Tracking with Siamese Track-RCNN | [pdf] | arxiv(2020) | Siamese Track-RCNN |
Online Single Stage Joint Detection and Tracking | [pdf] | CVPR2020 | RetinaTrack |
A Simple Baseline for Multi-Object Tracking | [pdf][code] | arXiv(2019) | FairMOT |
Tracking Objects as Points | [pdf] [code] | arXiv(2019) | CenterTrack |
Refinements in Motion and Appearance for Online Multi-Object Tracking | [pdf] [code] | arXiv(2019) | MIFT |
Multiple Object Tracking by Flowing and Fusing | [pdf] | arXiv(2019) | FFT |
A Unified Object Motion and Affinity Model for Online Multi-Object Tracking | [pdf][code] | CVPR2020 | UMA |
DeepMOT:A Differentiable Framework for Training Multiple Object Trackers | [pdf] [code] | CVPR2020 | DeepMOT |
Online multiple pedestrian tracking using deep temporal appearance matching association | [pdf] [code] | arXiv(2019) | DD_TAMA19 |
Spatial-temporal relation networks for multi-object tracking | [pdf] | ICCV2019 | STRN |
Towards Real-Time Multi-Object Tracking | [pdf] [code] | arXiv(2019) | JDE(private) |
Multi-object tracking with multiple cues and switcher-aware classification | [pdf] | arXiv(2019) | LSST |
FAMNet: Joint learning of feature, affinity and multi-dimensional assignment for online multiple object tracking | [pdf] | ICCV2019 | FAMNet |
Online multi-object tracking with instance-aware tracker and dynamic model refreshment | [pdf] | WACV2019 | KCF |
Tracking without bells and whistles | [pdf] [code] | ICCV2019 | Tracktor |
MOTS: Multi-Object Tracking and Segmentation | [pdf] [code] | CVPR2019 | Track R-CNN |
Eliminating Exposure Bias and Metric Mismatch in Multiple Object Tracking | [pdf] [code] | CVPR2019 | SAS_MOT17 |
Deep affinity network for multiple object tracking | [pdf] [code] | PAMI(2019) | DAN |
Recurrent autoregressive networks for online multi-object tracking | [pdf] | WACV2018 | RAN |
Real-time multiple people tracking with deeply learned candidate selection and person re-identification | [pdf] [code] | ICME2018 | MOTDT |
Online multi-object tracking with dual matching attention networks | [pdf] [code] | ECCV2018 | DMAN |
Extending IOU Based Multi-Object Tracking by Visual Information | [pdf] [code] | AVSS2018 | V-IOU |
Online Multi-target Tracking using Recurrent Neural Networks | [pdf] [code] | AAAI2017 | MOT-RNN |
Detect to Track and Track to Detect | [pdf] [code] | ICCV2017 | D&T(private) |
Online multi-object tracking using CNN-based single object tracker with spatial-temporal attention mechanism | [pdf] | ICCV2017 | STAM |
Tracking the untrackable: Learning to track multiple cues with long-term dependencies | [pdf] | ICCV2017 | AMIR |
Simple online and realtime tracking with a deep association metric | [pdf] [code] | ICIP2017 | DeepSort |
High-speed tracking-by-detection without using image information | [pdf] [code] | AVSS2017 | IOU Tracker |
Simple online and realtime tracking | [pdf] [code] | ICIP2016 | Sort |
Temporal dynamic appearance modeling for online multi-person tracking | [pdf] | CVIU(2016) | TDAM |
Online multi-object tracking via structural constraint event aggregation | [pdf] | CVPR2016 | SCEA |
Online Multi-Object Tracking Via Robust Collaborative Model and Sample Selection | [pdf] [code] | CVIU2016 | RCMSS |
Learning to Track: Online Multi-Object Tracking by Decision Making | [pdf] [code] | ICCV2015 | MDP |
Learning to Divide and Conquer for Online Multi-Target Tracking | [pdf] [code] | ICCV2015 | LDCT |
Robust online multi-object tracking based on tracklet confidence and online discriminative appearance learning | [pdf] [code] | CVPR2014 | CMOT |
The Way They Move: Tracking Targets with Similar Appearance | [pdf] [code] | ICCV2013 | SMOT |
Online Multi-Person Tracking by Tracker Hierarchy | [pdf] [code] | AVSS2012 | OMPTTH |
离线跟踪(Batch)
Name | Source | Publication | Notes |
---|---|---|---|
Lifted Disjoint Paths with Application in Multiple Object Tracking | [pdf] [code] | ICML2020 | Lif_T |
Learning non-uniform hypergraph for multi-object tracking | [pdf] | AAAI2019 | NT |
Learning a Neural Solver for Multiple Object Tracking | [pdf] [code] | CVPR2020 | MPNTracker |
Deep learning of graph matching | [pdf] | CVPR2018 | 深度图匹配 |
muSSP: Efficient Min-cost Flow Algorithm for Multi-object Tracking | [pdf] [code] | NIPS(2019) | muSSP |
Exploit the connectivity: Multi-object tracking with trackletnet | [pdf] [code] | ACM mm 2019 | TNT(eTC) |
Multiple people tracking using body and joint detections | [pdf] | CVPRW2019 | JBNOT |
Aggregate Tracklet Appearance Features for Multi-Object Tracking | [pdf] | SPL(2019) | NOTA |
Customized multi-person tracker | [pdf] | ACCV2018 | HCC |
Multi-object tracking with neural gating using bilinear lstm | [pdf] | ECCV2018 | MHT_bLSTM |
Trajectory Factory: Tracklet Cleaving and Re-connection by Deep Siamese Bi-GRU for Multiple Object Tracking | [pdf] | ICME2018 | GCRE |
Multiple People Tracking with Lifted Multicut and Person Re-identification | [pdf] | CVPR2017 | LMP |
Deep network flow for multi-object tracking | [pdf] | CVPR2017 | - |
Non-markovian globally consistent multi-object tracking | [pdf] [code] | ICCV2017 | - |
Multi-Object Tracking with Quadruplet Convolutional Neural Networks | [pdf] | CVPR2017 | Quad-CNN |
Enhancing detection model for multiple hypothesis tracking | [pdf] | CVPRW2017 | EDMT |
POI: Multiple Object Tracking with High Performance Detection and Appearance Feature | [pdf] | ECCV2016 | KNDT |
Multiple hypothesis tracking revisited | [pdf] [code] | ICCV2015 | MHT-DAM |
Near-Online Multi-target Tracking with Aggregated Local Flow Descriptor | [pdf] | ICCV2015 | NOMT |
On Pairwise Costs for Network Flow Multi-Object Tracking | [pdf] [code] | CVPR2015 | - |
Multiple Target Tracking Based on Undirected Hierarchical Relation Hypergraph | [pdf] [code] | CVPR2014 | H2T |
Continuous Energy Minimization for Multi-Target Tracking | [pdf] [code] | CVPR2014 | CEM |
GMCP-Tracker: Global Multi-object Tracking Using Generalized Minimum Clique Graphs | [pdf] [code] | ECCV2012 | GMCP |
Multiple Object Tracking using K-Shortest Paths Optimization | [pdf] [code] | PAMI2011 | KSP |
Global data association for multi-object tracking using network flows | [pdf] [code] | CVPR2008 | - |
跨摄像头跟踪(MTMC)
Name | Source | Publication | Notes |
---|---|---|---|
Locality Aware Appearance Metric for Multi-Target Multi-Camera Tracking | [pdf] code | CVPR2019 Workshop | LAAM |
CityFlow: A City-Scale Benchmark for Multi-Target Multi-Camera Vehicle Tracking and Re-Identification | [pdf] | CVPR2019 | CityFlow |
Features for multi-target multi-camera tracking and re-identification | [pdf] [code] | CVPR2018 | DeepCC(MTMC) |
Rolling Shutter and Radial Distortion Are Features for High Frame Rate Multi-Camera Tracking | [pdf] | CVPR2018 | - |
Towards a Principled Integration of Multi-Camera Re-Identification andTracking through Optimal Bayes Filters | [pdf] [code] | CVPR2017 | towards-reid-tracking |
3D&多模态跟踪
Name | Source | Publication | Notes |
---|---|---|---|
Joint 3D Tracking and Forecasting with Graph Neural Network and Diversity Sampling | [pdf] [code] | arxiv | GNNTrkForecast |
Graph Neural Network for 3D Multi-Object Tracking with Multi-Feature Learning | [pdf] [code] | CVPR2020 | GNN3DMOT |
Robust Multi-Modality Multi-Object Tracking | [pdf] [code] | ICCV2019 | mmMOT |
A baseline for 3D Multi-Object Tracking | [pdf] [code] | arXiv | - |
综述
Multiple Object Tracking: A Literature Review
Machine Learning Methods for Solving Assignment Problems in Multi-Target Tracking
Deep Learning in Video Multi-Object Tracking_ A Survey
Globally-Optimal Greedy Algorithms for Tracking a Variable Number of Objects
数据集
MOT:包含2D MOT2015、3D MOT2015、MOT16、MOT17和MOT17Det等多个子数据集,提供了ACF、DPM、Faster RCNN、SDP等多个检测器输入。包含不同的相机视角、相机运动、场景和时间变化以及密集场景。
KITTI:提供了汽车和行人的标注,场景较稀疏。
TUD Stadtmitte:包含3D人体姿态识别、多视角行人检测和朝向检测、以及行人跟踪的标注,相机视角很低,数据集不大。
ETHZ:由手机拍摄的多人跟踪数据集,包含三个场景。
EPFL:多摄像头采集的行人检测和跟踪数据集,每隔摄像头离地2米,实验人员就是一个实验室的,分为实验室、校园、平台、通道、篮球场这5个场景,每个场景下都有多个摄像头,每个摄像头拍摄2分钟左右。
KIT AIS:空中拍摄的,只有行人的头
PETS:比较早期的视频,有各式各样的行人运动。
DukeMTMC:多摄像头多行人跟踪。
MOTS:多目标跟踪与分割。