/MO-VLN-mindspore

Primary LanguagePythonApache License 2.0Apache-2.0

MO-VLN: A Multi-Task Benchmark for Open-set Zero-Shot Vision-and-Language Navigation

This is a MindSpore implementation of the paper.

Installing Dependencies

pip install https://ms-release.obs.cn-north-4.myhuaweicloud.com/2.0.0/MindSpore/unified/x86_64/mindspore-2.0.0-cp37-cp37m-linux_x86_64.whl --trusted-host ms-release.obs.cn-north-4.myhuaweicloud.com -i https://pypi.tuna.tsinghua.edu.cn/simple

git clone -b dev https://gitee.com/mindspore/mindformers.git
cd mindformers
bash build.sh

Setup

Clone the repository and install other requirements:

git clone https://github.com/liangcici/MO-VLN.git
cd MO-VLN/
pip install -r requirements.txt

Setting up dataset

  • Generate data for ObjectNav (goal-conditioned navigation given a specific object category).
python data_preprocess/gen_objectnav.py --map_id 3

map_id indicates specific scene: {3: Starbucks; 4: TG; 5: NursingRoom}.

Usage

Run models with FBE:

  • For ObjectNav:
python zero_shot_eval.py --sem_seg_model_type clip --map_id 3

Citation

@article{liang2023mo,
  title={MO-VLN: A Multi-Task Benchmark for Open-set Zero-Shot Vision-and-Language Navigation},
  author={Liang, Xiwen and Ma, Liang and Guo, Shanshan and Han, Jianhua and Xu, Hang and Ma, Shikui and Liang, Xiaodan},
  journal={arXiv preprint arXiv:2306.10322},
  year={2023}
}