apple2373 / action-general.icdl

Action Recognition based on Cross-Situational Action-object Statistics

@inproceedings{tsutsui2022action,
	author = {Satoshi Tsutsui and Xizi Wang and Guangyuan Weng and Yayun Zhang and Chen Yu and David Crandall},
	booktitle = {{International Conference on Development and Learning (ICDL)}},
	title = {Action Recognition based on Cross-Situational Action-object Statistics},
	year = {2022}
}

• https://arxiv.org/abs/2208.07344

Environment

• pytorch 1.3.1
• scikit-video 1.1.11 see environment.yml for deatils.

conda env create --prefix=conda -f environment.yml
conda activate ./conda

How to run

• set up something-something dataset on ./data/something

python main.py -g 0

e.g.,

python main.py -g 0 \
--train ./data/cvpr-v4/exp_N-375_NumCommonNouns-0_NumUniqueNouns-1_seed-0_missingexcluded_samplingseed-0.train.csv \
--val ./data/cvpr-v4/exp_N-375_NumCommonNouns-0_NumUniqueNouns-1_seed-0_missingexcluded_samplingseed-0.val.csv \
--test ./data/cvpr-v4/exp_N-375_NumCommonNouns-0_NumUniqueNouns-1_seed-0_missingexcluded_samplingseed-0.test.csv

Experiment logs

• https://drive.google.com/file/d/1AM2VcSvpfS32vIn6f0lvSNQxOKSLRznB/edit
• Download it, then tar xf experiments.tar
• results-main.ipynb has the plots used in the paper.

Note

• Currently this repository only has the 3D-ResNet18 model for RGB.
• The data splits are available in ./data/cvpr-v4.