Pose Estimation On Crowdpose

Metriken

AP

AP50

AP75

APM

Test

Ergebnisse

Leistungsergebnisse verschiedener Modelle zu diesem Benchmark

						Paper Title
BUCTD-W48 (w/cond. input from PETR, and generative sampling)	78.5	-	-	-	-	Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity
ViTPose-G	78.3	85.3	81.4	86.6	-	ViTPose: Simple Vision Transformer Baselines for Human Pose Estimation
BUCTD-W48 (w/cond. input from PETR)	76.7	-	-	-	-	Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity
SwinV2-L 1K-MIM	75.5	-	-	-	-	Revealing the Dark Secrets of Masked Image Modeling
SwinV2-B 1K-MIM	74.9	-	-	-	-	Revealing the Dark Secrets of Masked Image Modeling
BUCTD-W48	72.9	-	-	-	-	Rethinking pose estimation in crowds: overcoming the detection information-bottleneck and ambiguity
OpenPifPaf	70.5	89.1	76.1	-	-	OpenPifPaf: Composite Fields for Semantic Keypoint Detection and Spatio-Temporal Association
MIPNet (HRNet-W48)	70.0	-	-	71.1	-	Multi-Instance Pose Networks: Rethinking Top-Down Pose Estimation
KAPAO-L	68.9	89.4	75.6	69.9	76.6	Rethinking Keypoint Representations: Modeling Keypoints and Poses as Objects for Multi-Person Human Pose Estimation
KAPAO-M	67.1	88.8	73.4	68.1	75.2	Rethinking Keypoint Representations: Modeling Keypoints and Poses as Objects for Multi-Person Human Pose Estimation
Hourglass-104	65.2	85.9	69.5	66.2	-	Greedy Offset-Guided Keypoint Grouping for Human Pose Estimation
KAPAO-S	63.8	87.7	69.4	64.8	72.1	Rethinking Keypoint Representations: Modeling Keypoints and Poses as Objects for Multi-Person Human Pose Estimation

0 of 12 row(s) selected.

Pose Estimation On Crowdpose | SOTA | HyperAI