zjdcts/st-gcn-pytorch

Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition

Human Action Recognition

This code is implemented based on Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition.
And this code only supports the Florence 3D action dataset.
Please download from here at ./dataset/

Environments

• python 3.6
• pytorch 1.0

Run

python main.py

File Details

File Name	Description
preprocess.py	Preprocess the Florence 3D action dataset. Each frame is unified into 32 frames, separated by train / valid / test and dropped into a file.
config.py	Hyperparameter and data path setting
main.py	Execution File. Loads data and models, and performs training and testing.
model.py	Defines the General Graph Convolutional Network (GGCN) class. Construct an adjacency matrix with three consecutive graphs, and call several layers.
layer.py	Defines the graph convolution, standard convolution and classifier layer.
metric.py	Defines the accuracy function