Semantic Binary Segmentation Using Convolutional Networks Without Decoders

Shubhra Aich, William van der Kamp, Ian Stavness

CVPRW 2018 pp. 197-201

doi:10.1109/CVPRW.2018.00032 /cvprw/2018/aich2018cvprw-semantic/

Abstract

In this paper, we propose an efficient architecture for semantic image segmentation using the depth-to-space (D2S) operation. Our D2S model is comprised of a standard CNN encoder followed by a depth-to-space reordering of the final convolutional feature maps. Our approach eliminates the decoder portion of traditional encoder-decoder segmentation models and reduces the amount of computation almost by half. As a participant of the DeepGlobe Road Extraction competition, we evaluate our models on the corresponding road segmentation dataset. Our highly efficient D2S models exhibit comparable performance to standard segmentation models with much lower computational cost.

PDF CVPRW Semantic Scholar

Cite

Text

Aich et al. "Semantic Binary Segmentation Using Convolutional Networks Without Decoders." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2018. doi:10.1109/CVPRW.2018.00032

Markdown

[Aich et al. "Semantic Binary Segmentation Using Convolutional Networks Without Decoders." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2018.](https://mlanthology.org/cvprw/2018/aich2018cvprw-semantic/) doi:10.1109/CVPRW.2018.00032

BibTeX

@inproceedings{aich2018cvprw-semantic,
  title     = {{Semantic Binary Segmentation Using Convolutional Networks Without Decoders}},
  author    = {Aich, Shubhra and van der Kamp, William and Stavness, Ian},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2018},
  pages     = {197-201},
  doi       = {10.1109/CVPRW.2018.00032},
  url       = {https://mlanthology.org/cvprw/2018/aich2018cvprw-semantic/}
}