By Zilong Huang, Chunyu Wang, Xinggang Wang, Wenyu Liu and Jingdong Wang.

This code is a implementation of the experiments in the paper Semantic Image Segmentation by Scale-Adaptive Networks, which is accepted by Transactions on Image Processing. The code is developed based on the Caffe framework.

SAN is released under the MIT License (refer to the LICENSE file for details).

If you find SANet useful in your research, please consider citing:

@article{huang2019sanet,
    title={Semantic Image Segmentation by Scale-Adaptive Networks},
    author={Huang, Zilong and Wang, Chunyu and Wang, Xinggang and Liu, Wenyu and Wang, Jingdong},
    journal={IEEE Transactions on Image Processing},
    year={2019},
    publisher={IEEE}
}

• caffe (deeplabv2 version): deeplabv2 caffe installation instructions are available at https://bitbucket.org/aquariusjay/deeplab-public-ver2. Note, you need to compile caffe with python wrapper and support for python layers. Then add the caffe python path into tools/findcaffe.py.

• Run:

      $ python tools/train.py --solver YOUR_SOLVER --weight IMAGENET_PRETRAINED_MODEL --gpu GPU_ID

The corresponding solver files and input image lists are put in config and list floders.

The work was mainly done during an internship at MSRA.