Neuromorphic high-frequency 3D dancing pose estimation in dynamic environment

• This repository is the official implementation of paper Neuromorphic high-frequency 3D dancing pose estimation in dynamic environment in Neurocomputing, 547, 2023.
• Here is the introduction video of the project and datasets we collected/generated.
• The dataset is released here on UCSD DataPlanet Website.

It is built with pytorch-lightning and miracleyoo's pl template, with the basic toolsets borrowed from lifting_events_to_3d_hpe.

Default Training Command Used: python main.py --data_dir=<DATA_DIR> --use_mask='unet' --mask_net_path=<MASK_NET_PATH.ckpt> --batch_size=16 --gpus=8 --partial_dataset=0.3