APNet: Urban-level Scene Segmentation of Aerial Images and Point Clouds

Paper | Poster | Slides

This repo is the official implementation for the ICCVW'23 paper: APNet: Urban-level Scene Segmentation of Aerial Images and Point Clouds

Dependencies

To run our code first install the dependencies with:

conda env create -f environment.yaml

Dataset

SensatUrban Pre-processing

The code will be released after cleaning.

Running the code

Then run the following command:

sh run_files/train_eval.sh

Citation

If you use this repo, please cite as :

@inproceedings{wei2023apnet,
    author = {Weijie Wei and Martin R. Oswald and Fatemeh Karimi Nejadasl and Theo Gevers},
    title = {{APNet: Urban-level Scene Segmentation of Aerial Images and Point Clouds}},
    booktitle = {{Proceedings of the IEEE International Conference on Computer Vision Workshops (ICCVW)}},
    year = {2023}
}

Acknowledgement

Our code is heavily inspired by the following projects:

1. RandLA-Net: https://github.com/QingyongHu/RandLA-Net
2. RandLA-Net-pytorch: https://github.com/tsunghan-wu/RandLA-Net-pytorch
3. HRNet: https://github.com/HRNet/HRNet-Semantic-Segmentation
4. KPConv: https://github.com/HuguesTHOMAS/KPConv-PyTorch
5. KPRNet: https://github.com/DeyvidKochanov-TomTom/kprnet
6. SensatUrban-BEV-Seg3D: https://github.com/zouzhenhong98/SensatUrban-BEV-Seg3D

Thanks for their contributions.