MotionAGFormer: Enhancing 3D Human Pose Estimation with a Transformer-GCNFormer Network

Benchmark Model Rank Results
3d-human-pose-estimation-on-mpi-inf-3dhpMotionAGFormer-L (T=81)#2MPJPE: 16.2AUC: 85.3PCK: 98.2
3d-human-pose-estimation-on-mpi-inf-3dhpMotionAGFormer-S (T=81)#4MPJPE: 17.1AUC: 84.5PCK: 98.3
3d-human-pose-estimation-on-mpi-inf-3dhpMotionAGFormer-B (T=81)#6MPJPE: 18.2AUC: 84.2PCK: 98.3
3d-human-pose-estimation-on-mpi-inf-3dhpMotionAGFormer-XS (T=27)#7MPJPE: 19.2AUC: 83.5PCK: 98.2
classification-on-full-body-parkinsonsMotionAGFormer#6F1-score (weighted): 0.42
monocular-3d-human-pose-estimation-on-human3MotionAGFormer-L#2Average MPJPE (mm): 38.4Use Video Sequence: YesFrames Needed: 243
monocular-3d-human-pose-estimation-on-human3MotionAGFormer-B#3Average MPJPE (mm): 38.4Use Video Sequence: YesFrames Needed: 243
monocular-3d-human-pose-estimation-on-human3MotionAGFormer-S#10Average MPJPE (mm): 42.5Use Video Sequence: YesFrames Needed: 81
monocular-3d-human-pose-estimation-on-human3MotionAGFormer-XS#15Average MPJPE (mm): 45.1Use Video Sequence: YesFrames Needed: 27