/BESNet

Primary LanguagePython

Boundary Enhancing Semantic Context Network for Parsing High-resolution Remote Sensing Images

  • Updating......

The code is based on FCN_8s

Datasets

  • ISPRS Vaihingen dataset
  • ISPRS Potsdam dataset
  • The original dataset can be requested for the download from here.
  • You should cut the training images as well as corresponding labels into patches with an overlap of 171 pixels.

Requiements

  • Python == 3.7.10
  • PyTorch == 1.8.1
  • CUDA ==10.1

Train

For Vaihingen, run:

sh train_vaihingen_fcn.sh

For Potsdam, run:

sh train_potsdam_fcn.sh

The results will be saved in the ./snapshots/ folder.

Test

For test, run:

sh Eval.sh

For different datasets, please manully change dataset in Eval.sh.