LEADERBOARD.md

December 3, 2020 · View on GitHub

Leaderboard

Unsupervised and domain adaptive re-ID methods on public benchmarks. To add some papers not included, you could create an issue or a pull request. Note: the following results are copied from their original papers.

For trained models by OpenUnReID, please refer to MODEL_ZOO.md.

Unsupervised learning (USL) on object re-ID
Unsupervised domain adaptation (UDA) on object re-ID

Unsupervised learning on object re-ID

Market-1501

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL+	NeurIPS'20	PyTorch (OpenUnReID)	76.0	89.5	96.2	97.5	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMT+	ICLR'20	PyTorch (OpenUnReID)	74.3	88.1	96.0	97.5	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
SpCL	NeurIPS'20	PyTorch	73.1	88.1	95.1	97.0	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
Strong Baseline	N/A	PyTorch (OpenUnReID)	70.5	87.9	95.7	97.1	N/A
HCT	CVPR'20	Empty	56.4	80.0	91.6	95.2	Hierarchical Clustering with Hard-batch Triplet Loss for Person Re-identification
MMCL	CVPR'20	PyTorch	45.5	80.3	89.4	92.3	Unsupervised Person Re-Identification via Multi-Label Classification
SSL	CVPR'20	PyTorch (Unofficial)	37.8	71.7	83.8	87.4	Unsupervised Person Re-identification via Softened Similarity Learning
BUC	AAAI'19	PyTorch	38.3	66.2	79.6	84.5	A Bottom-up Clustering Approach to Unsupervised Person Re-identification

DukeMTMC-reID

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL+	NeurIPS'20	PyTorch (OpenUnReID)	67.1	82.4	90.8	93.0	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMT+	ICLR'20	PyTorch (OpenUnReID)	60.3	75.6	86.0	89.2	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
SpCL	NeurIPS'20	PyTorch	65.3	81.2	90.3	92.2	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
Strong Baseline	N/A	PyTorch (OpenUnReID)	54.7	72.9	83.5	87.2	N/A
HCT	CVPR'20	Empty	50.7	69.6	83.4	87.4	Hierarchical Clustering with Hard-batch Triplet Loss for Person Re-identification
MMCL	CVPR'20	PyTorch	40.2	65.2	75.9	80.0	Unsupervised Person Re-Identification via Multi-Label Classification
SSL	CVPR'20	PyTorch (Unofficial)	28.6	52.5	63.5	68.9	Unsupervised Person Re-identification via Softened Similarity Learning
BUC	AAAI'19	PyTorch	27.5	47.4	62.6	68.4	A Bottom-up Clustering Approach to Unsupervised Person Re-identification

MSMT17

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL	NeurIPS'20	PyTorch	19.1	42.3	55.6	61.2	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMCL	CVPR'20	PyTorch	11.2	35.4	44.8	49.8	Unsupervised Person Re-Identification via Multi-Label Classification

Unsupervised domain adaptation on object re-ID

Market-1501 -> DukeMTMC-reID

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL+	NeurIPS'20	PyTorch (OpenUnReID)	70.4	83.8	91.2	93.4	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMT+	ICLR'20	PyTorch (OpenUnReID)	67.7	80.3	89.9	92.9	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
SpCL	NeurIPS'20	PyTorch	68.8	82.9	90.1	92.5	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MEB-Net	ECCV'20	PyTorch	66.1	79.6	88.3	92.2	Multiple Expert Brainstorming for Domain Adaptive Person Re-identification
MMT	ICLR'20	PyTorch	65.1	78.0	88.8	92.5	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
Strong Baseline	N/A	PyTorch (OpenUnReID)	60.4	75.9	86.2	89.8	N/A
AD-Cluster	CVPR'20	PyTorch	54.1	72.6	82.5	85.5	AD-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-identification
SNR	CVPR'20	-	58.1	76.3	-	-	Style Normalization and Restitution for Generalizable Person Re-identification
MMCL	CVPR'20	PyTorch	51.4	72.4	82.9	85.0	Unsupervised Person Re-Identification via Multi-Label Classification
ECN++	TPAMI'20	-	54.4	74.0	83.7	87.4	Learning to Adapt Invariance in Memory for Person Re-identification
UDA_TP	PR'20	PyTorch or OpenUnReID	49.0	68.4	80.1	83.5	Unsupervised Domain Adaptive Re-Identification: Theory and Practice
SSG	ICCV'19	PyTorch	53.4	73.0	80.6	83.2	Self-similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-identification
PCB-PAST	ICCV'19	PyTorch	54.3	72.4	-	-	Self-Training With Progressive Augmentation for Unsupervised Cross-Domain Person Re-Identification
CR-GAN	ICCV'19	-	48.6	68.9	80.2	84.7	Instance-Guided Context Rendering for Cross-Domain Person Re-Identification
PDA-Net	ICCV'19	-	45.1	63.2	77.0	82.5	Cross-Dataset Person Re-Identification via Unsupervised Pose Disentanglement and Adaptation
UCDA	ICCV'19	-	31.0	47.7	-	-	A Novel Unsupervised Camera-aware Domain Adaptation Framework for Person Re-identification
ECN	CVPR'19	PyTorch	40.4	63.3	75.8	80.4	Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-identification
HHL	ECCV'18	PyTorch	33.4	60.2	73.9	79.5	Generalizing A Person Retrieval Model Hetero- and Homogeneously
SPGAN	CVPR'18	PyTorch or OpenUnReID	22.3	41.1	56.6	63.0	Image-Image Domain Adaptation with Preserved Self-Similarity and Domain-Dissimilarity for Person Re-identification
TJ-AIDL	CVPR'18	-	23.0	44.3	59.6	65.0	Transferable Joint Attribute-Identity Deep Learning for Unsupervised Person Re-Identification
PUL	TOMM'18	PyTorch	16.4	30.0	43.4	48.5	Unsupervised Person Re-identification: Clustering and Fine-tuning

DukeMTMC-reID -> Market-1501

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL+	NeurIPS'20	PyTorch (OpenUnReID)	78.2	90.5	96.6	97.8	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMT+	ICLR'20	PyTorch (OpenUnReID)	80.9	92.2	97.6	98.4	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
SpCL	NeurIPS'20	PyTorch	76.7	90.3	96.2	97.7	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MEB-Net	ECCV'20	PyTorch	76.0	89.9	96.0	97.5	Multiple Expert Brainstorming for Domain Adaptive Person Re-identification
Strong Baseline	N/A	PyTorch (OpenUnReID)	75.6	90.9	96.6	97.8	N/A
MMT	ICLR'20	PyTorch	71.2	87.7	94.9	96.9	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
AD-Cluster	CVPR'20	PyTorch	68.3	86.7	94.4	96.5	AD-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-identification
SNR	CVPR'20	-	61.7	82.8	-	-	Style Normalization and Restitution for Generalizable Person Re-identification
MMCL	CVPR'20	PyTorch	60.4	84.4	92.8	95.0	Unsupervised Person Re-Identification via Multi-Label Classification
ECN++	TPAMI'20	-	63.8	84.1	92.8	95.4	Learning to Adapt Invariance in Memory for Person Re-identification
UDA_TP	PR'20	PyTorch or OpenUnReID	53.7	75.8	89.5	93.2	Unsupervised Domain Adaptive Re-Identification: Theory and Practice
SSG	ICCV'19	PyTorch	58.3	80.0	90.0	92.4	Self-similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-identification
PCB-PAST	ICCV'19	PyTorch	54.6	78.4	-	-	Self-Training With Progressive Augmentation for Unsupervised Cross-Domain Person Re-Identification
CR-GAN	ICCV'19	-	54.0	77.7	89.7	92.7	Instance-Guided Context Rendering for Cross-Domain Person Re-Identification
PDA-Net	ICCV'19	-	47.6	75.2	86.3	90.2	Cross-Dataset Person Re-Identification via Unsupervised Pose Disentanglement and Adaptation
UCDA	ICCV'19	-	30.9	60.4	-	-	A Novel Unsupervised Camera-aware Domain Adaptation Framework for Person Re-identification
ECN	CVPR'19	PyTorch	43.0	75.1	87.6	91.6	Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-identification
HHL	ECCV'18	PyTorch	31.4	62.2	78.8	84.05	Generalizing A Person Retrieval Model Hetero- and Homogeneously
SPGAN	CVPR'18	PyTorch or OpenUnReID	22.8	51.5	70.1	76.8	Image-Image Domain Adaptation with Preserved Self-Similarity and Domain-Dissimilarity for Person Re-identification
TJ-AIDL	CVPR'18	-	26.5	58.2	74.8	81.1	Transferable Joint Attribute-Identity Deep Learning for Unsupervised Person Re-Identification
PUL	TOMM'18	PyTorch	20.5	45.5	60.7	66.7	Unsupervised Person Re-identification: Clustering and Fine-tuning

Market-1501 -> MSMT17

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL	NeurIPS'20	PyTorch	26.8	53.7	65.0	69.8	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMT	ICLR'20	PyTorch	22.9	49.2	63.1	68.8	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
MMCL	CVPR'20	PyTorch	15.1	40.8	51.8	56.7	Unsupervised Person Re-Identification via Multi-Label Classification
ECN++	TPAMI'20	-	15.2	40.4	53.1	58.7	Learning to Adapt Invariance in Memory for Person Re-identification
SSG	ICCV'19	PyTorch	13.2	31.6	-	49.6	Self-similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-identification
ECN	CVPR'19	PyTorch	8.5	25.3	36.3	42.1	Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-identification
PTGAN	CVPR'18	-	2.9	10.2	-	24.4	Person Transfer GAN to Bridge Domain Gap for Person Re-Identification

DukeMTMC-reID -> MSMT17

Method	Venue	Code	mAP(%)	R@1(%)	R@5(%)	R@10(%)	Reference
SpCL	NeurIPS'20	PyTorch	26.5	53.1	65.8	70.5	Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
MMT	ICLR'20	PyTorch	23.3	50.1	63.9	69.8	Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
MMCL	CVPR'20	PyTorch	16.2	43.6	54.3	58.9	Unsupervised Person Re-Identification via Multi-Label Classification
ECN++	TPAMI'20	-	16.0	42.5	55.9	61.5	Learning to Adapt Invariance in Memory for Person Re-identification
SSG	ICCV'19	PyTorch	13.3	32.2	-	51.2	Self-similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-identification
ECN	CVPR'19	PyTorch	10.2	30.2	41.5	46.8	Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-identification
PTGAN	CVPR'18	-	3.3	11.8	-	27.4	Person Transfer GAN to Bridge Domain Gap for Person Re-Identification