3D Human Pose Estimation On 3Dpw

评估指标

Acceleration Error

MPJPE

MPVPE

PA-MPJPE

评测结果

各个模型在此基准测试上的表现结果

模型名称	Acceleration Error	MPJPE	MPVPE	PA-MPJPE	Paper Title	Repository
TePose (T=6)	11.4	84.6	100.3	52.3	Live Stream Temporally Embedded 3D Human Body Pose and Shape Estimation
TCMR (T=16 w/o 3DPW)	7	95	111.5	55.8	Beyond Static Features for Temporally Consistent 3D Human Pose and Shape from a Video
Self-Attentive	77.9	85.8	100.6	50.4	Self-Attentive 3D Human Pose and Shape Estimation from Videos	-
DSR	-	85.7	99.5	51.7	Learning to Regress Bodies from Images using Differentiable Semantic Rendering
Cross Dataset Generalization	-	89.7	-	65.2	Predicting Camera Viewpoint Improves Cross-dataset Generalization for 3D Human Pose Estimation	-
XFormer (HRNet)	-	75	87.1	45.7	XFormer: Fast and Accurate Monocular 3D Body Capture	-
TAR (N=9)	7.7	62.7	74.4	40.6	Temporal-Aware Refinement for Video-based Human Pose and Shape Recovery	-
ZeDO (S=1,J=17)	-	69.7	-	40.3	Back to Optimization: Diffusion-based Zero-Shot 3D Human Pose Estimation
Hand4Whole	-	-	-	54.4	Accurate 3D Hand Pose Estimation for Whole-Body 3D Human Mesh Estimation
TP-Net	-	-	-	92.2	Learning 3D Human Pose from Structure and Motion
WHAM (ViT)	-	57.8	68.7	35.9	WHAM: Reconstructing World-grounded Humans with Accurate 3D Motion
DeciWatch-PARE	-	75.5	-	46.4	DeciWatch: A Simple Baseline for 10x Efficient 2D and 3D Pose Estimation
VirtualMarker	-	67.5	77.9	41.3	3D Human Mesh Estimation from Virtual Markers
SPIN	-	96.9	116.4	59.2	Learning to Reconstruct 3D Human Pose and Shape via Model-fitting in the Loop
GTRS	-	88.5	106.2	58.9	A Lightweight Graph Transformer Network for Human Mesh Reconstruction from 2D Human Pose
INT-2 (ResNet-50)	16.5	75.6	87.9	42	Capturing the motion of every joint: 3D human pose and shape estimation with independent tokens
3DHR-Co (w/ 2D GT)	-	63.72	-	42.11	3DHR-Co: A Collaborative Test-time Refinement Framework for In-the-Wild 3D Human-Body Reconstruction Task	-
ProHMR	-	-	-	65	Probabilistic Modeling for Human Mesh Recovery
GLoT	6.6	80.7	96.3	50.6	Global-to-Local Modeling for Video-based 3D Human Pose and Shape Estimation
ZeDO (Cross Dataset)	-	80.9	-	42.6	Back to Optimization: Diffusion-based Zero-Shot 3D Human Pose Estimation

0 of 142 row(s) selected.