A Simple but Effective Downsampling Module For Semantic Segmentation
-
install pytorch_wavelets
git clone https://github.com/fbcotter/pytorch_wavelets cd pytorch_wavelets pip install .
Please refer for more details: https://github.com/fbcotter/pytorch_wavelets -
install semantic segmentation package
pip install segmentation-models-pytorch
Please refer for more details: https://github.com/qubvel/segmentation_models.pytorch -
You can replace the downsample method, like maxpooling, averagepooling, convolution with stride, with our HWD directly.
-
Dataset (1) Camvid: https://github.com/alexgkendall/SegNet-Tutorial (2) Synapse: https://pan.baidu.com/s/1PoCxb_BNAalm37pmPTOBmw Code: asdf
Please cite our paper if it is helpful to you. Thanks.
@article{XU2023109819, title = {Haar Wavelet Downsampling: A Simple but Effective Downsampling Module for Semantic Segmentation}, journal = {Pattern Recognition}, pages = {109819}, year = {2023}, issn = {0031-3203}, doi = {https://doi.org/10.1016/j.patcog.2023.109819}, url = {https://www.sciencedirect.com/science/article/pii/S0031320323005174}, author = {Guoping Xu and Wentao Liao and Xuan Zhang and Chang Li and Xinwei He and Xinglong Wu}, keywords = {Semantic segmentation, Downsampling, Haar wavelet, Information Entropy} }