STGCN++ is a variant of STGCN we developed in PYSKL with some modifications in the architecture of the spatial module and the temporal module. We provide STGCN++ trained on NTURGB+D with 2D skeletons (HRNet) and 3D skeletons with PYSKL training setting. We provide checkpoints for four modalities: Joint, Bone, Joint Motion, and Bone Motion. The architecture of STGCN++ is described in PYSKL tech report.
@misc{duan2022PYSKL,
url = {https://arxiv.org/abs/2205.09443},
author = {Duan, Haodong and Wang, Jiaqi and Chen, Kai and Lin, Dahua},
title = {PYSKL: Towards Good Practices for Skeleton Action Recognition},
publisher = {arXiv},
year = {2022}
}
We release numerous checkpoints trained with various modalities, annotations on NTURGB+D and NTURGB+D 120. The accuracy of each modality links to the weight file.
Note
- We use the linear-scaling learning rate (Initial LR ∝ Batch Size). If you change the training batch size, remember to change the initial LR proportionally.
- For Two-Stream results, we adopt the 1 (Joint):1 (Bone) fusion. For Four-Stream results, we adopt the 2 (Joint):2 (Bone):1 (Joint Motion):1 (Bone Motion) fusion.
You can use the following command to train a model.
bash tools/dist_train.sh ${CONFIG_FILE} ${NUM_GPUS} [optional arguments]
# For example: train STGCN++ on NTURGB+D XSub (3D skeleton, Joint Modality) with 8 GPUs, with validation, and test the last and the best (with best validation metric) checkpoint.
bash tools/dist_train.sh configs/stgcn++/stgcn++_ntu60_xsub_3dkp/j.py 8 --validate --test-last --test-best
You can use the following command to test a model.
bash tools/dist_test.sh ${CONFIG_FILE} ${CHECKPOINT_FILE} ${NUM_GPUS} [optional arguments]
# For example: test STGCN++ on NTURGB+D XSub (3D skeleton, Joint Modality) with metrics `top_k_accuracy`, and dump the result to `result.pkl`.
bash tools/dist_test.sh configs/stgcn++/stgcn++_ntu60_xsub_3dkp/j.py checkpoints/SOME_CHECKPOINT.pth 8 --eval top_k_accuracy --out result.pkl