📄️ 🤖 Create SMPLH Fit
This section guides you through creating a fit between SMPL+H human body parameters and your robot's skeleton. This fitting process is crucial for transferring human motion data to robot controls, establishing the correspondence between human joints and robot links.
📄️ 🚅 Train Poseformer for Retargeting
This section introduces how to train an efficient retargeting network using the SMPL+H model created in the previous section on the AMASS dataset. The trained network will support real-time motion retargeting at over 1000 FPS for arbitrary motions.
📄️ 🔀 Convert AMASS Actions
This section describes how to convert the entire AMASS dataset into robot-feasible, learnable data for imitation learning. This process combines the SMPL+H fitting and PoseFormer training from previous sections to create a comprehensive dataset ready for reinforcement learning environments.