ICENet (Indoor Cognition Extraction Network)

1. Installation of dependent libraries

    pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

2.Dataset Download

Click on Indoor Space Cognition Dataset to download the dataset.

password: fengbuxi

3. Task-level configuration file usage

• Configure global_config.py to add the new task's category mapping dictionary directly when adding a new task, and change the value of pretrained_task at the same time.

• The base model of the last mission should be placed under ~/.torch/models/ folder, the system will read the base model automatically, the default name is mtcnet.pth, if you change the name, please synchronize to change the pth_path variable under train.py.

4.Train

python train.py

5.Eval

python eval.py

6.Demo

python demo.py

Connection

+ Email: [email protected]

+ WebSite: https://www.jaingjianwu.cn/

Citation

@article{ICENet,
  title={Scene-level spatial cognition information extraction method for indoor safety monitoring of elderly-assisting robots}, 
  author={Jianwu Jiang and Jiadi Chen and Jingwen Li and Xiao Dong and Qingyang Wang and Bo Song},
  journal={Pending},
  year={2024}
}

Acknowledgements

• This model is constructed based on PSPNet.

• This model is built on the open source framework awesome-semantic-segmentation-pytorch.