This repo contains code for our NeurIPS Datasets and Benchmarks 2022 paper:

EgoTaskQA: Understanding Human Tasks in Egocentric Videos

Baoxiong Jia, Ting Lei, Song-Chun Zhu, Siyuan Huang

For data download, please check our website for instructions and details.

We provide all environment configurations in requirements.txt. In our experiments, we used NVIDIA CUDA 11.3 on Ubuntu 20.04 and need this additional step for version control on pytorch:

$ conda create -n egotaskqa python=3.8
$ pip install -r requirements.txt
$ pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu113 

Similar CUDA version should also be acceptable with corresponding version control for torch and torchvision. We refer the authors to Generation and Experiment for details on quetsion-answer generation, balancing, data split, and baseline experiments. For these two functionalities, please checkout the corresponding sub-directory for code and instructions.

If you find our paper and/or code helpful, please consider citing:

@inproceedings{jia2022egotaskqa,
    title = {EgoTaskQA: Understanding Human Tasks in Egocentric Videos},
    author = {Jia, Baoxiong and Lei, Ting and Zhu, Song-Chun and Huang, Siyuan},
    booktitle = {The 36th Conference on Neural Information Processing Systems (NeurIPS 2022) Track on Datasets and Benchmarks},
    year = {2022}
}

We thank all colleagues from VCLA and BIGAI for fruitful discussions. We would also like to thank the anonymous reviewers for their constructive feedback.