We release the PyTorch code of LSTA
Please cite our paper if you find the repo and the paper useful.
@InProceedings{Sudhakaran_2019_CVPR,
author = {Sudhakaran, Swathikiran and Escalera, Sergio and Lanz, Oswald},
title = {{LSTA: Long Short-Term Attention for Egocentric Action Recognition}},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2019}
}
- Python 3.5
- Pytorch 0.3.1
To train the models, run the script train_rgb.sh, which contains:
python main_rgb.py --dataset gtea_61 --root_dir dataset --outDir experiments --stage 1 \
--seqLen 25 --trainBatchSize 32 --numEpochs 200 --lr 0.001 --stepSize 25 75 150 \
--decayRate 0.1 --memSize 512 --outPoolSize 100 --evalInterval 5 --split 2
Testing on the trained models can be done by running the script test_rgb.sh
The pre-trained models can be downloaded from the following Google Drive link
- EPIC-KITCHENS code
- Flow and two stream codes
- Pre-trained models