zw20yz / STRNet

PyTorch implementation for the [paper]https://doi.org/10.1007/s11633-021-1289-9

STRNet

PyTorch implementation for the [paper]https://doi.org/10.1007/s11633-021-1289-9

STRNet: Triple-Stream Spatial Temporal Relation Network for Action Recognition[J]. In-ternational Journal of Automation and Computing, 2021,1-13.

Environment:

• Python 3.5.2
• PyTorch 0.4.1
• TorchVison: 0.2.1

Clone this repo

git clone https://github.com/xzwwx/STRNet

Data Preparation

Please refer to TSN repo and TSM repo for the detailed guide of data pre-processing.

Training

• To train STRNet on Something-Something V1 dataset with 8 frames:

bash ./run2.sh

Pre-trained Models

Currently, we do not provide the original pre-trained models on Kinetics, since we have reorganized the structure of the codes and renamed the modules of STRNet for public release. The old models cannot be loaded with new names. We plan to retrain the models with the new codes and release the models for evaluation.