dim_block1, dim_block2, dim_block3, respectively to set block dimension where the output channels are equal to 64, 128 and 256.To set the block dimensions of the windowed version of Temporal Transformer:
In order to run T-TR-agcn and ST-TR-agcn configurations, please set agcn: True. Python3 ensemble.py Adaptive Configuration (AGCN) An element in position (i, j) represents the correlation between joint i and joint j, resulting from self-attention. The heatmaps are 25 x 25 matrices, where each row and each column represents a body joint. Visualizations of Spatial Transformer logits Skeleton-based action recognition via spatial and temporal transformer networks, Chiara Plizzari, Marco Cannici, Matteo Matteucci, Computer Vision and Image Understanding, Volumes 208-209, 2021, 103219, ISSN 1077-3142, CVIU ICPR International Workshops and Challenges, 2021, Proceedings Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition, Chiara Plizzari, Marco Cannici, Matteo Matteucci, Pattern Recognition. Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition, Chiara Plizzari, Marco Cannici, Matteo Matteucci, ArXiv This repository contains the implementation of the model presented in the following paper: Spatial Temporal Transformer Network Introduction