Enhanced 3D Human Pose Estimation from Videos by using Attention-Based Neural Network with Dilated Convolutions

Benchmark Model Rank Results
3d-human-pose-estimation-on-human36mAttention (T=243 CPN)Average MPJPE (mm): 44.8Using 2D ground-truth joints: No
3d-human-pose-estimation-on-humaneva-iAttention (T=27 MA)Mean Reconstruction Error (mm): 15.4