Multi Person Pose Estimation On Crowdpose

评估指标

AP Easy

AP Hard

AP Medium

mAP @0.5:0.95

评测结果

各个模型在此基准测试上的表现结果

					Paper Title
RTMO-l	88.8	77.2	84.7	83.8	RTMO: Towards High-Performance One-Stage Real-Time Multi-Person Pose Estimation
BUCTD-W48 (w/cond. input from PETR, and generative sampling)	83.9	72.3	79.0	78.5	Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity
I²R-Net (1st stage: HRFormer-B)	83.8	69.3	78.1	77.4	I^2R-Net: Intra- and Inter-Human Relation Network for Multi-Person Pose Estimation
ED-Pose (Swin-L)	83.0	68.3	77.3	76.6	Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation
HRFormer-B	80.0	62.4	73.5	72.4	HRFormer: High-Resolution Transformer for Dense Prediction
BAPose (W32)	79.9	61.3	73.4	72.2	BAPose: Bottom-Up Pose Estimation with Disentangled Waterfall Representations
TransPose-H	79.5	62.2	72.9	71.8	TransPose: Keypoint Localization via Transformer
SCIO (HRNet-48)	-	-	72.2	71.5	Self-Constrained Inference Optimization on Structural Groups for Human Pose Estimation
HigherHRNet (ScaleNet_P4)	-	-	-	71.3	ScaleNAS: One-Shot Learning of Scale-Aware Representations for Visual Recognition
MIPNet (HRNet-W48)	78.1	59.4	71.1	70.0	Multi-Instance Pose Networks: Rethinking Top-Down Pose Estimation
CenterGroup	76.6	61.5	70.0	69.4	The Center of Attention: Center-Keypoint Grouping via Attention for Multi-Person Pose Estimation
HigherHRNet(HR-Net-48)	75.8	58.9	68.1	67.6	HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation
Joint-candidate SPPE +	75.5	57.4	66.3	66.0	CrowdPose: Efficient Crowded Scenes Pose Estimation and A New Benchmark
OccNet	75.2	53.1	66.6	65.5	Human Pose Estimation for Real-World Crowded Scenarios
Hourglass-104	73.8	54.8	66.2	65.2	Greedy Offset-Guided Keypoint Grouping for Human Pose Estimation
SPM	70.3	55.7	64.5	63.7	Single-Stage Multi-Person Pose Machines
AlphaPose	71.2	51.1	61.4	61.0	RMPE: Regional Multi-person Pose Estimation
Simple baseline	71.4	51.2	61.2	60.8	Simple Baselines for Human Pose Estimation and Tracking
ROMP+CAR	-	-	-	58.6	Monocular, One-stage, Regression of Multiple 3D People
LitePose-S	-	-	-	58.3	Lite Pose: Efficient Architecture Design for 2D Human Pose Estimation

0 of 23 row(s) selected.

Command Palette

Multi Person Pose Estimation On Crowdpose

评估指标

评测结果

Command Palette

Multi Person Pose Estimation On Crowdpose

评估指标

评测结果