Home

Awesome

3D Skeleton Based Person Re-Identification (SRID)

Awesome PRs Welcome Visits Badge

A professionally curated list of resources (papers, codes, data, etc.) on 3D Skeleton Based Person Re-ID (SRID), which is the first work to comprehensively and systematically summarize the recent advances of SRID research to the best of our knowledge.

We will continuously update this list with the latest resources. Should you find any missed resources (papers/codes) or errors, please feel free to open an issue or contribute a pull request.

For more papers and resources on Skeleton-Based Models (Action Recognition, Pose Estimation, etc.) from top-tier AI conferences and journals, kindly refer to This Repo.

Survey Paper

A Survey on 3D Skeleton Based Person Re-Identification: Approaches, Designs, Challenges, and Future Directions

By Haocong Rao and Chunyan Miao.

Taxonomy of SRID

image

Archives and Resources

<!-- vscode-markdown-toc --> <!-- vscode-markdown-toc-config numbering=true autoSave=true /vscode-markdown-toc-config --> <!-- /vscode-markdown-toc -->

Benchmark Datasets

Overview of commonly-used benchmark datasets for 3D skeleton-based person re-identification and their statistics. The number of skeletons in the training set is estimated and reported. “Ego” denotes a single or egocentric view. We also include person re-ID datasets with 2D/3D skeletons estimated from RGB videos.

# DatasetsYearSource# ID# Skeleton# View
PAVIS RGBD-ID2012Kinect V179Ego
BIWI RGBD-ID2013Kinect V150205.8KEgo
IAS-Lab RGBD-ID2013Kinect V11189.0KEgo
KGBD2014Kinect V1164188.7KEgo
KinectREID2015Kinect V1714.8K7
UPCV12015Kinect V13013.1KEgo
UPCV22016Kinect V23026.3KEgo
Florence 3D Re-ID2016Kinect V21618.0KEgo
KS202017Kinect V22036.0K5
Freestyle Walks2017Kinect V290Ego
CAISA-B-3D2020Estimated from RGB videos124706.5K11
OUMVLP-Pose-2D2020Estimated from RGB videos103076667.0K14
PoseTrackReID-2D2020Estimated from RGB videos535053.6K

Studies by Different Categories

image

Hand-Crafted Methods

Sequence Learning Methods

Graph Learning Methods

Studies by Different Years

2023

2022

2021

2020

<!-- - [Human skeleton mutual learning for person re-identification](https://doi.org/10.1016/j.neucom.2019.12.120) (_Neurocomputing 2020_) -->

2019

<!-- - [SKEPRID: Pose and Illumination Change-Resistant Skeleton-Based Person Re-Identification](https://dl.acm.org/doi/pdf/10.1145/3243217) (_ACM Transactions on Multimedia Computing, Communications, and Applications 2019_) -->

2018

2017

2016

2015

2014

Before 2014

Leaderboards

The results are mainly from the paper of (CVPR 2023) [paper] (TranSG: Transformer-Based Skeleton Graph Prototype Contrastive Learning with Structure-Trajectory Prompted Reconstruction for Person Re-Identification).

MethodsBIWI-SBIWI-WKS20
mAPRank-1Rank-5Rank-10mAPRank-1Rank-5Rank-10mAPRank-1Rank-5Rank-10
Hand-Crafted Methods
${D_{\text{PG}}}$ (Pattern Recognition 2020)6.718.545.463.88.76.515.520.311.335.261.570.5
${D_{13}}$ (Person Re-Identification 2014)13.128.353.165.917.214.220.623.718.939.471.781.7
${D_{16}}$ (Computers&Graphics 2019)16.732.655.768.318.817.025.329.624.051.777.186.9
Sequence Learning Methods
PoseGait (PR 2020)9.914.040.756.711.18.823.031.223.549.480.990.2
AGE (IJCAI 2020)8.925.143.161.612.611.721.427.38.943.270.180.0
SGELA (TPAMI 2021)15.125.851.864.419.011.714.014.721.245.065.075.1
SimMC (IJCAI 2022)12.341.766.676.819.924.536.744.522.366.480.787.0
Hi-MPC (IJCV 2023)17.447.570.378.622.627.340.348.822.069.683.587.1
Graph Learning Methods
MG-SCR (IJCAI 2021)7.620.146.964.111.910.820.329.410.446.375.484.0
SM-SGE (ACM MM 2021)10.131.356.369.115.213.225.833.59.545.971.981.2
SPC-MGR (Arxiv 2022)16.034.157.369.819.418.931.540.521.759.079.086.2
TranSG (CVPR 2023)30.168.786.591.826.932.744.952.246.273.686.390.2
MethodsIAS-AIAS-BKGBD
mAPRank-1Rank-5Rank-10mAPRank-1Rank-5Rank-10mAPRank-1Rank-5Rank-10
Hand-Crafted Methods
${D_{\text{PG}}}$ (Pattern Recognition 2020)11.016.439.553.410.616.041.257.32.130.049.158.1
${D_{13}}$ (Person Re-Identification 2014)24.540.058.767.623.743.768.676.71.917.034.444.2
${D_{16}}$ (Computers&Graphics 2019)25.242.762.970.724.544.569.180.24.031.250.959.8
Sequence Learning Methods
PoseGait (PR 2020)17.528.455.769.220.828.951.662.913.950.667.072.6
AGE(IJCAI 2020)13.431.154.867.412.831.152.364.20.92.95.67.5
SGELA (TPAMI 2021)13.216.730.244.014.022.240.850.24.538.153.560.0
SimMC (IJCAI 2022)18.744.865.372.922.946.368.177.011.754.966.270.6
Hi-MPC (IJCV 2023)23.245.667.375.425.348.270.277.810.256.970.275.1
Graph Learning Methods
MG-SCR (IJCAI 2021)14.136.459.669.512.932.456.569.46.944.058.764.6
SM-SGE (ACM MM 2021)13.634.060.571.613.338.964.175.84.438.254.260.7
SPC-MGR (Arxiv 2022)24.241.966.375.624.143.368.479.46.940.857.565.0
TranSG (CVPR 2023)32.849.268.576.239.459.177.087.020.259.073.178.2

Citation

If you found this paper/repository useful, please consider citing:

@article{rao2024survey,
  title={A Survey on 3D Skeleton Based Person Re-Identification: Approaches, Designs, Challenges, and Future Directions},
  author={Rao, Haocong and Miao, Chunyan},
  journal={arXiv preprint arXiv:2401.15296},
  year={2024}
}