There are two steps to generate the datasets:
Step1:
git clone git@github.com:johnny09/GMR.git
cd GMR
python scripts/bvh_to_robot_npz_dataset.py
--src_folder <path_to_lafan1_bvh_file_folder> \
--tgt_folder <path_to_output_npz_file_folder> \
--format lafan1 \
--robot pnd_adam_sp \
--target_fps 30
The generated file is in .
The info in the generated motion data including:
"fps";
"root_pos";
"root_rot";
"dof_pos";
"dof_vel";
Step2:
This step mainly do two things:
1.convert "dof_pos" and "dof_vel" in the motion data from step1 to "qpos" and "qvel"
2.interpolate the motion data to the target_fps
3.extend the motion data
git clone git@github.com:johnny09/loco-mujoco.git
cd loco-mujoco
python convert_custom_format_and_extend_new.py \
-b \
-i <path_to_input_motion_data_folder> \
-o <path_to_output_motion_data_folder> \
-of 50 \
--env PndAdamSp