ML Research Wiki / Benchmarks / Multi-Person Pose Estimation / CrowdPose

CrowdPose

Multi-Person Pose Estimation Benchmark

Performance Over Time

📊 Showing 28 results | 📏 Metric: mAP @0.5:0.95

Top Performing Models

Rank	Model	Paper	mAP @0.5:0.95	Date	Code
1	RTMO-l 📚	RTMO: Towards High-Performance One-Stage Real-Time Multi-Person Pose Estimation	83.80	2023-12-12	📦 open-mmlab/mmpose
2	BUCTD-W48 (w/cond. input from PETR, and generative sampling)	Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity	78.50	2023-06-13	📦 amathislab/BUCTD
3	I²R-Net (1st stage: HRFormer-B)	I^2R-Net: Intra- and Inter-Human Relation Network for Multi-Person Pose Estimation	77.40	2022-06-22	📦 leijue222/Intra-and-Inter-Human-Relation-Network-for-MPEE
4	ED-Pose (Swin-L) 📚	Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation	76.60	2023-02-03	📦 idea-research/ed-pose 📦 michel-liu/grouppose 📦 michel-liu/grouppose-paddle
5	DETRPose-X	DETRPose: Real-time end-to-end transformer model for multi-person pose estimation	75.10	2025-06-16	📦 SebastianJanampa/DETRPose
6	DETRPose-L	DETRPose: Real-time end-to-end transformer model for multi-person pose estimation	73.30	2025-06-16	📦 SebastianJanampa/DETRPose
7	HRFormer-B	HRFormer: High-Resolution Transformer for Dense Prediction	72.40	2021-10-18	📦 HRNet/HRFormer
8	BAPose (W32)	BAPose: Bottom-Up Pose Estimation with Disentangled Waterfall Representations	72.20	2021-12-20	📦 bmartacho/BAPose
9	DETRPose-M	DETRPose: Real-time end-to-end transformer model for multi-person pose estimation	72.00	2025-06-16	📦 SebastianJanampa/DETRPose
10	TransPose-H	TransPose: Keypoint Localization via Transformer	71.80	2020-12-28	📦 yangsenius/TransPose

All Papers (28)

RTMO: Towards High-Performance One-Stage Real-Time Multi-Person Pose Estimation

2023

RTMO-l

open-mmlab/mmpose

Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity

2023

BUCTD-W48 (w/cond. input from PETR, and generative sampling)

amathislab/BUCTD

I^2R-Net: Intra- and Inter-Human Relation Network for Multi-Person Pose Estimation

2022

I²R-Net (1st stage: HRFormer-B)

leijue222/Intra-and-Inter-Human-Relation-Network-for-MPEE

Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation

2023

ED-Pose (Swin-L)

idea-research/ed-pose michel-liu/grouppose michel-liu/grouppose-paddle

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

2025

DETRPose-X

SebastianJanampa/DETRPose

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

2025

DETRPose-L

SebastianJanampa/DETRPose

HRFormer: High-Resolution Transformer for Dense Prediction

2021

HRFormer-B

HRNet/HRFormer

BAPose: Bottom-Up Pose Estimation with Disentangled Waterfall Representations

2021

BAPose (W32)

bmartacho/BAPose

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

2025

DETRPose-M

SebastianJanampa/DETRPose

TransPose: Keypoint Localization via Transformer

2020

TransPose-H

yangsenius/TransPose

Self-Constrained Inference Optimization on Structural Groups for Human Pose Estimation

2022

SCIO (HRNet-48)

ScaleNAS: One-Shot Learning of Scale-Aware Representations for Visual Recognition

2020

HigherHRNet (ScaleNet_P4)

Multi-Instance Pose Networks: Rethinking Top-Down Pose Estimation

2021

MIPNet (HRNet-W48)

rawalkhirodkar/MIPNet

The Center of Attention: Center-Keypoint Grouping via Attention for Multi-Person Pose Estimation

2021

CenterGroup

dvl-tum/center-group

HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation

2019

HigherHRNet(HR-Net-48)

PaddlePaddle/PaddleDetection open-mmlab/mmpose

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

2025

DETRPose-S

SebastianJanampa/DETRPose

CrowdPose: Efficient Crowded Scenes Pose Estimation and A New Benchmark

2018

Joint-candidate SPPE +

open-mmlab/mmpose jeffffffli/CrowdPose

Human Pose Estimation for Real-World Crowded Scenarios

2019

OccNet

thomasgolda/Human-Pose-Estimation-for-Real-World-Crowded-Scenarios

Greedy Offset-Guided Keypoint Grouping for Human Pose Estimation

2021

Hourglass-104

hellojialee/OffsetGuided

Single-Stage Multi-Person Pose Machines

2019

SPM

murdockhou/Single-Stage-Multi-person-Pose-Machines

OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields

2018

OpenPose

CMU-Perceptual-Computing-Lab/openpose chonyy/AI-basketball-analysis

RMPE: Regional Multi-person Pose Estimation

2016

AlphaPose

MVIG-SJTU/AlphaPose osmr/imgclsmob

Simple Baselines for Human Pose Estimation and Tracking

2018

Simple baseline

PaddlePaddle/PaddleDetection open-mmlab/mmpose

Monocular, One-stage, Regression of Multiple 3D People

2020

ROMP+CAR

Arthur151/ROMP cai-jianfeng/ROMP_mindspore

Lite Pose: Efficient Architecture Design for 2D Human Pose Estimation

2022

LitePose-S

mit-han-lab/litepose

Mask R-CNN

2017

Mask R-CNN

tensorflow/models facebookresearch/detectron2

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

2025

DETRPose-N

SebastianJanampa/DETRPose

Monocular, One-stage, Regression of Multiple 3D People

2020

ROMP

Arthur151/ROMP cai-jianfeng/ROMP_mindspore

Model	Paper	mAP @0.5:0.95	Date
RTMO-l	RTMO: Towards High-Performance One-Stage Real-Tim…	83.80	2023-12-12
BUCTD-W48 (w/cond. input from PETR, and generative sampling)	Rethinking pose estimation in crowds: overcoming …	78.50	2023-06-13
I²R-Net (1st stage: HRFormer-B)	I^2R-Net: Intra- and Inter-Human Relation Network…	77.40	2022-06-22
ED-Pose (Swin-L)	Explicit Box Detection Unifies End-to-End Multi-P…	76.60	2023-02-03
DETRPose-X	DETRPose: Real-time end-to-end transformer model …	75.10	2025-06-16
DETRPose-L	DETRPose: Real-time end-to-end transformer model …	73.30	2025-06-16
HRFormer-B	HRFormer: High-Resolution Transformer for Dense P…	72.40	2021-10-18
BAPose (W32)	BAPose: Bottom-Up Pose Estimation with Disentangl…	72.20	2021-12-20
DETRPose-M	DETRPose: Real-time end-to-end transformer model …	72.00	2025-06-16
TransPose-H	TransPose: Keypoint Localization via Transformer	71.80	2020-12-28
SCIO (HRNet-48)	Self-Constrained Inference Optimization on Struct…	71.50	2022-07-06
HigherHRNet (ScaleNet_P4)	ScaleNAS: One-Shot Learning of Scale-Aware Repres…	71.30	2020-11-30
MIPNet (HRNet-W48)	Multi-Instance Pose Networks: Rethinking Top-Down…	70.00	2021-01-27
CenterGroup	The Center of Attention: Center-Keypoint Grouping…	69.40	2021-10-11
HigherHRNet(HR-Net-48)	HigherHRNet: Scale-Aware Representation Learning …	67.60	2019-08-27
DETRPose-S	DETRPose: Real-time end-to-end transformer model …	67.40	2025-06-16
Joint-candidate SPPE +	CrowdPose: Efficient Crowded Scenes Pose Estimati…	66.00	2018-12-02
OccNet	Human Pose Estimation for Real-World Crowded Scen…	65.50	2019-07-16
Hourglass-104	Greedy Offset-Guided Keypoint Grouping for Human …	65.20	2021-07-07
SPM	Single-Stage Multi-Person Pose Machines	63.70	2019-08-24
OpenPose	OpenPose: Realtime Multi-Person 2D Pose Estimatio…	62.70	2018-12-18
AlphaPose	RMPE: Regional Multi-person Pose Estimation	61.00	2016-12-01
Simple baseline	Simple Baselines for Human Pose Estimation and Tr…	60.80	2018-04-17
ROMP+CAR	Monocular, One-stage, Regression of Multiple 3D P…	58.60	2020-08-27
LitePose-S	Lite Pose: Efficient Architecture Design for 2D H…	58.30	2022-05-03
Mask R-CNN	Mask R-CNN	57.20	2017-03-20
DETRPose-N	DETRPose: Real-time end-to-end transformer model …	56.00	2025-06-16
ROMP	Monocular, One-stage, Regression of Multiple 3D P…	55.60	2020-08-27

CrowdPose

Performance Over Time

Edit Benchmark Results

Edit Result

Top Performing Models

All Papers (28)

RTMO: Towards High-Performance One-Stage Real-Time Multi-Person Pose Estimation

Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity

I^2R-Net: Intra- and Inter-Human Relation Network for Multi-Person Pose Estimation

Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

HRFormer: High-Resolution Transformer for Dense Prediction

BAPose: Bottom-Up Pose Estimation with Disentangled Waterfall Representations

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

TransPose: Keypoint Localization via Transformer

Self-Constrained Inference Optimization on Structural Groups for Human Pose Estimation

ScaleNAS: One-Shot Learning of Scale-Aware Representations for Visual Recognition

Multi-Instance Pose Networks: Rethinking Top-Down Pose Estimation

The Center of Attention: Center-Keypoint Grouping via Attention for Multi-Person Pose Estimation

HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

CrowdPose: Efficient Crowded Scenes Pose Estimation and A New Benchmark

Human Pose Estimation for Real-World Crowded Scenarios

Greedy Offset-Guided Keypoint Grouping for Human Pose Estimation

Single-Stage Multi-Person Pose Machines

OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields

RMPE: Regional Multi-person Pose Estimation

Simple Baselines for Human Pose Estimation and Tracking

Monocular, One-stage, Regression of Multiple 3D People

Lite Pose: Efficient Architecture Design for 2D Human Pose Estimation

Mask R-CNN

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

Monocular, One-stage, Regression of Multiple 3D People