Kali-Hac / 3D-skeleton-based-person-re-ID-survey

A professionally curated list of resources (paper, code, data, etc.) on 3D Skeleton Based Person Re-ID.

3D Skeleton Based Person Re-Identification (SRID)

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

Archives and Resources

• Benchmark Datasets
• Studies by Different Categories
  • Hand-Crafted Methods
  • Sequence Learning Methods
  • Graph Learning Methods
• Studies by Different Years
  • 2023
  • 2022
  • 2021
  • 2020
  • 2019
  • 2018
  • 2017
  • 2016
  • 2015
  • 2014
  • Before 2014
• Leaderboards

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.

# Datasets	Year	Source	# ID	# Skeleton	# View
PAVIS RGBD-ID	2012	Kinect V1	79	—	Ego
BIWI RGBD-ID	2013	Kinect V1	50	205.8K	Ego
IAS-Lab RGBD-ID	2013	Kinect V1	11	89.0K	Ego
KGBD	2014	Kinect V1	164	188.7K	Ego
KinectREID	2015	Kinect V1	71	4.8K	7
UPCV1	2015	Kinect V1	30	13.1K	Ego
UPCV2	2016	Kinect V2	30	26.3K	Ego
Florence 3D Re-ID	2016	Kinect V2	16	18.0K	Ego
KS20	2017	Kinect V2	20	36.0K	5
Freestyle Walks	2017	Kinect V2	90	—	Ego
CAISA-B-3D	2020	Estimated from RGB videos	124	706.5K	11
OUMVLP-Pose-2D	2020	Estimated from RGB videos	10307	6667.0K	14
PoseTrackReID-2D	2020	Estimated from RGB videos	5350	53.6K	—

Studies by Different Categories

Hand-Crafted Methods

• One-Shot Person Re-identification with a Consumer Depth Camera (Person Re-Identification 2014)

• 3D reconstruction of freely moving persons for re-identification with a depth sensor (ICRA 2014)

• A feature-based approach to people re-identification using skeleton keypoints (ICRA 2014)

• People re-identification using 3D descriptor with skeleton information (International Conference on Informatics, Electronics & Vision 2015)

• Multimodal person reidentification using RGB-D cameras (IEEE Transactions on Circuits and Systems for Video Technology 2015)

• Person Identification Using Anthropometric and Gait Data from Kinect Sensor (AAAI 2015)

• Long term person re-identification from depth cameras using facial and skeleton data (ICPR 2016)

• Human identification from freestyle walks using posture-based gait feature (IEEE Transactions on Information Forensics and Security 2017)

• Context-aware person re-identification in the wild via fusion of gait and anthropometric features (International Conference on Automatic Face & Gesture Recognition 2017)

• Cross-context analysis for long-term view-point invariant person re-identification via soft-biometrics using depth sensor (VISIGRAPP 2018)

• Enhanced skeleton and face 3D data for person re-identification from depth cameras (Computers&Graphics 2019)

• Person Re-Identification from Different Views based on Dynamic Linear Combination of Distances (Multimedia Tools and Applications 2021) [Github]

• Re-visiting k-Reciprocal Distance Re-ranking for Skeleton-Based Person Re-identification (IEEE Signal Processing Letters 2022)

Sequence Learning Methods

• Person identification by walking gesture using skeleton sequences (Advanced Concepts for Intelligent Vision Systems 2020)

• Learning 3D spatiotemporal gait feature by convolutional network for person identification (Neurocomputing 2020)

• A model-based gait recognition method with body pose and human prior knowledge (Pattern Recognition 2020) [Github] (PoseGait is extended for SRID in recent studies from 2021-2023)

• Self-Supervised Gait Encoding with Locality-Aware Attention for Person Re-Identification (IJCAI 2020) [Github]

• A Self-Supervised Gait Encoding Approach with Locality-Awareness for 3D Skeleton Based Person Re-Identification (IEEE Transactions on Pattern Analysis and Machine Intelligence 2021) [Github]

• Human Identification System Using 3D Skeleton-Based Gait Features and LSTM Model (Journal of Visual Communication and Image Representation 2022)

• SimMC: Simple Masked Contrastive Learning of Skeleton Representations for Unsupervised Person Re-Identification (IJCAI 2022) [Github]

• Hierarchical Skeleton Meta-prototype Contrastive Learning with Hard Skeleton Mining for Unsupervised Person Re-identification (International Journal of Computer Vision 2023) [Github]

Graph Learning Methods

• SM-SGE: A Self-Supervised Multi-Scale Skeleton Graph Encoding Framework for Person Re-Identification (ACM MM 2021) [Github]

• Multi-Level Graph Encoding with Structural-Collaborative Relation Learning for Skeleton-Based Person Re-Identification (IJCAI 2021) [Github]

• Skeleton Prototype Contrastive Learning with Multi-level Graph Relation Modeling for Unsupervised Person Re-identification (Arxiv (Preprint) 2022) [Github]

• TranSG: Transformer-based Skeleton Graph Prototype Contrastive Learning with Structure-Trajectory Prompted Reconstruction for Person Re-identification (CVPR 2023) [Github]

Studies by Different Years

2023

• TranSG: Transformer-based Skeleton Graph Prototype Contrastive Learning with Structure-Trajectory Prompted Reconstruction for Person Re-identification (CVPR 2023) [Github]

• Hierarchical Skeleton Meta-prototype Contrastive Learning with Hard Skeleton Mining for Unsupervised Person Re-identification (International Journal of Computer Vision 2023) [Github]

2022

• SimMC: Simple Masked Contrastive Learning of Skeleton Representations for Unsupervised Person Re-Identification (IJCAI 2022) [Github]

• Re-visiting k-Reciprocal Distance Re-ranking for Skeleton-Based Person Re-identification (IEEE Signal Processing Letters 2022)

• Skeleton Prototype Contrastive Learning with Multi-level Graph Relation Modeling for Unsupervised Person Re-identification (Arxiv (Preprint) 2022) [Github]

• Human Identification System Using 3D Skeleton-Based Gait Features and LSTM Model (Journal of Visual Communication and Image Representation 2022)

2021

• SM-SGE: A Self-Supervised Multi-Scale Skeleton Graph Encoding Framework for Person Re-Identification (ACM MM 2021) [Github]

• Multi-Level Graph Encoding with Structural-Collaborative Relation Learning for Skeleton-Based Person Re-Identification (IJCAI 2021) [Github]

• A Self-Supervised Gait Encoding Approach with Locality-Awareness for 3D Skeleton Based Person Re-Identification (IEEE Transactions on Pattern Analysis and Machine Intelligence 2021) [Github]

• Person Re-Identification from Different Views based on Dynamic Linear Combination of Distances (Multimedia Tools and Applications 2021) [Github]

2020

• Self-Supervised Gait Encoding with Locality-Aware Attention for Person Re-Identification (IJCAI 2020) [Github]

• Learning 3D spatiotemporal gait feature by convolutional network for person identification (Neurocomputing 2020)

• A model-based gait recognition method with body pose and human prior knowledge (Pattern Recognition 2020) [Github] (PoseGait is extended for SRID in recent studies from 2021-2023)

• Person identification by walking gesture using skeleton sequences (Advanced Concepts for Intelligent Vision Systems 2020)

2019

• Enhanced skeleton and face 3D data for person re-identification from depth cameras (Computers&Graphics 2019)

2018

• Cross-context analysis for long-term view-point invariant person re-identification via soft-biometrics using depth sensor (VISIGRAPP 2018)

2017

• Human identification from freestyle walks using posture-based gait feature (IEEE Transactions on Information Forensics and Security 2017)

• Context-aware person re-identification in the wild via fusion of gait and anthropometric features (International Conference on Automatic Face & Gesture Recognition 2017)

2016

• Long term person re-identification from depth cameras using facial and skeleton data (ICPR 2016)

2015

• Person Identification Using Anthropometric and Gait Data from Kinect Sensor (AAAI 2015)

• People re-identification using 3D descriptor with skeleton information (International Conference on Informatics, Electronics & Vision 2015)

• Multimodal person reidentification using RGB-D cameras (IEEE Transactions on Circuits and Systems for Video Technology 2015)

2014

• 3D reconstruction of freely moving persons for re-identification with a depth sensor (ICRA 2014)

• A feature-based approach to people re-identification using skeleton keypoints (ICRA 2014)

• One-Shot Person Re-identification with a Consumer Depth Camera (Person Re-Identification 2014)

Before 2014

• Re-identification with RGB-D Sensors (ECCV Workshop 2012)

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).

Methods	BIWI-S				BIWI-W				KS20
	mAP	Rank-1	Rank-5	Rank-10	mAP	Rank-1	Rank-5	Rank-10	mAP	Rank-1	Rank-5	Rank-10
Hand-Crafted Methods
${D_{\text{PG}}}$ (Pattern Recognition 2020)	6.7	18.5	45.4	63.8	8.7	6.5	15.5	20.3	11.3	35.2	61.5	70.5
${D_{13}}$ (Person Re-Identification 2014)	13.1	28.3	53.1	65.9	17.2	14.2	20.6	23.7	18.9	39.4	71.7	81.7
${D_{16}}$ (Computers&Graphics 2019)	16.7	32.6	55.7	68.3	18.8	17.0	25.3	29.6	24.0	51.7	77.1	86.9
Sequence Learning Methods
PoseGait (PR 2020)	9.9	14.0	40.7	56.7	11.1	8.8	23.0	31.2	23.5	49.4	80.9	90.2
AGE (IJCAI 2020)	8.9	25.1	43.1	61.6	12.6	11.7	21.4	27.3	8.9	43.2	70.1	80.0
SGELA (TPAMI 2021)	15.1	25.8	51.8	64.4	19.0	11.7	14.0	14.7	21.2	45.0	65.0	75.1
SimMC (IJCAI 2022)	12.3	41.7	66.6	76.8	19.9	24.5	36.7	44.5	22.3	66.4	80.7	87.0
Hi-MPC (IJCV 2023)	17.4	47.5	70.3	78.6	22.6	27.3	40.3	48.8	22.0	69.6	83.5	87.1
Graph Learning Methods
MG-SCR (IJCAI 2021)	7.6	20.1	46.9	64.1	11.9	10.8	20.3	29.4	10.4	46.3	75.4	84.0
SM-SGE (ACM MM 2021)	10.1	31.3	56.3	69.1	15.2	13.2	25.8	33.5	9.5	45.9	71.9	81.2
SPC-MGR (Arxiv 2022)	16.0	34.1	57.3	69.8	19.4	18.9	31.5	40.5	21.7	59.0	79.0	86.2
TranSG (CVPR 2023)	30.1	68.7	86.5	91.8	26.9	32.7	44.9	52.2	46.2	73.6	86.3	90.2

Methods	IAS-A				IAS-B				KGBD
	mAP	Rank-1	Rank-5	Rank-10	mAP	Rank-1	Rank-5	Rank-10	mAP	Rank-1	Rank-5	Rank-10
Hand-Crafted Methods
${D_{\text{PG}}}$ (Pattern Recognition 2020)	11.0	16.4	39.5	53.4	10.6	16.0	41.2	57.3	2.1	30.0	49.1	58.1
${D_{13}}$ (Person Re-Identification 2014)	24.5	40.0	58.7	67.6	23.7	43.7	68.6	76.7	1.9	17.0	34.4	44.2
${D_{16}}$ (Computers&Graphics 2019)	25.2	42.7	62.9	70.7	24.5	44.5	69.1	80.2	4.0	31.2	50.9	59.8
Sequence Learning Methods
PoseGait (PR 2020)	17.5	28.4	55.7	69.2	20.8	28.9	51.6	62.9	13.9	50.6	67.0	72.6
AGE(IJCAI 2020)	13.4	31.1	54.8	67.4	12.8	31.1	52.3	64.2	0.9	2.9	5.6	7.5
SGELA (TPAMI 2021)	13.2	16.7	30.2	44.0	14.0	22.2	40.8	50.2	4.5	38.1	53.5	60.0
SimMC (IJCAI 2022)	18.7	44.8	65.3	72.9	22.9	46.3	68.1	77.0	11.7	54.9	66.2	70.6
Hi-MPC (IJCV 2023)	23.2	45.6	67.3	75.4	25.3	48.2	70.2	77.8	10.2	56.9	70.2	75.1
Graph Learning Methods
MG-SCR (IJCAI 2021)	14.1	36.4	59.6	69.5	12.9	32.4	56.5	69.4	6.9	44.0	58.7	64.6
SM-SGE (ACM MM 2021)	13.6	34.0	60.5	71.6	13.3	38.9	64.1	75.8	4.4	38.2	54.2	60.7
SPC-MGR (Arxiv 2022)	24.2	41.9	66.3	75.6	24.1	43.3	68.4	79.4	6.9	40.8	57.5	65.0
TranSG (CVPR 2023)	32.8	49.2	68.5	76.2	39.4	59.1	77.0	87.0	20.2	59.0	73.1	78.2