The official implementation of ECCV22 paper "Learning Spatial-Preserved Skeleton Representations for Few-Shot Action Recognition"
#environment
#environment
pytorch=1.7
python=3.8.10
tqdm=4.54.1
pynvml=8.0.4
#dataset
#dataset
Due to the oversize dataset（ beyond maximum limitation 100M, please download the public NTU RGB+D 120 dataset or use our subset when the work is accepted.
#run
#run
run train.py with full model(DASTM (full), 5-way-1-shot, STGCN, on NTU RGB+D 120).
