[ICIP 2022] Official implementation of the paper "Entropy-Based Feature Extraction For Real-Time Semantic Segmentation".

Pretrained RTEffNet+EPE model is available from Google Drive. For the testing of the model using the Cityscapes validation set run the following command:

python test.py --data_path path/to/validation/set --cuda --weights path/to/downloaded/weight

To train the RTEffNet+EPE module on Cityscapes dataset run the following command:

python main.py --dataset cityscapes --data_path ../../datasets/cityscapes/ --cuda

If you find the code useful for your research, please consider citing our works

@article{abrahamyanepe,
  title={Entropy-Based Feature Extraction For Real-Time Semantic Segmentation},
  author={Lusine, Abrahamyan and Nikos, Deligiannis},
  journal={Proceedings of the IEEE International Conference on Image Processing (ICIP)},
  publisher = {IEEE},
  year={2022}
}

Code for EfficientNet is borrowed from EfficientNet-PyTorch and for the Cityscapes loader from TORCHVISION.DATASETS and pytorch-semantic-segmentation.