Skip to content

Latest commit

 

History

History
95 lines (76 loc) · 3.49 KB

README.md

File metadata and controls

95 lines (76 loc) · 3.49 KB
Raw

AAAI 2025

TCPFormer: Learning Temporal Correlation with Implicit Pose Proxy for 3D Human Pose Estimation

Environment

The project is developed under the following environment:

  • Python 3.10.x
  • PyTorch 2.2.1
  • CUDA 12.1 For installation of the project dependencies, please run:
pip install -r requirements.txt

Dataset

Human3.6M

Preprocessing

  1. We follow the previous state-of-the-art method MotionBERT for dataset setup. Download the MotionBERT's preprocessed H3.6M data here and unzip it to 'data/motion3d'.
  2. Slice the motion clips by running the following python code in data/preprocess directory:

For our model with T = 243:

python h36m.py  --n-frames 243

or T = 81

python h36m.py  --n-frames 81

or T = 27

python h36m.py  --n-frames 81

MPI-INF-3DHP

Preprocessing

Please refer to P-STMO for dataset setup. After preprocessing, the generated .npz files (data_train_3dhp.npz and data_test_3dhp.npz) should be located at data/motion3d directory.

Training

After dataset preparation, you can train the model as follows:

Human3.6M

You can train Human3.6M with the following command:

python train.py --config <PATH-TO-CONFIG>

where config files are located at configs/h36m.

python train.py --config configs/h36m/TCPFormer_h36m_243.yaml

MPI-INF-3DHP

You can train MPI-INF-3DHP with the following command:

python train_3dhp.py --config <PATH-TO-CONFIG>

where config files are located at configs/mpi.

python train_3dhp.py --config configs/mpi/TCPFormer_mpi_81.yaml

Evaluation

Dataset frames Checkpoint
Human3.6M 81 download
Human3.6M 243 download
MPI-INF-3DHP 9 download
MPI-INF-3DHP 27 download
MPI-INF-3DHP 81 download

After downloading the weight from table above, you can evaluate Human3.6M models by:

python train.py --eval-only --checkpoint <CHECKPOINT-DIRECTORY> --checkpoint-file <CHECKPOINT-FILE-NAME> --config <PATH-TO-CONFIG>

For example if TCPFormer with T = 243 of H.36M is downloaded and put in checkpoint directory, then you can run:

python train.py --eval-only  --checkpoint checkpoint --checkpoint-file TCPFormer_h36m_243_379.pth.tr --config configs/h36m/TCPFormer_h36m_243.yaml

Similarly, TCPFormer with T = 81 of H.36M is downloaded and put in checkpoint directory, then you can run:

python train.py --eval-only  --checkpoint checkpoint --checkpoint-file TCPFormer_h36m_81_405.pth.tr --config configs/h36m/TCPFormer_h36m_81.yaml

For MPI-INF-3DHP dataset, you can download the checkpoint with T = 81 and put in checkpoint_mpi directory, then you can run:

python train_3dhp.py --eval-only  --checkpoint checkpoint_mpi --checkpoint-file TCPFormer_mpi_81.pth.tr --config configs/mpi/TCPFormer_mpi_81.yaml