/geoseg

A Pytorch Module for Building Segmentation and Outline Extraction

Primary LanguagePython

Geoseg - A Computer Vision Package for Automatic Building Segmentation and Outline extraction

Structure of directories

sub directories

Geoseg:
  ├── dataset/
  │   └── train, validate and test dataset
  ├── logs/
  │   ├── learning curve, logging, statistic, etc.
  ├── models/
  │   ├── fcn, fpn, u-net, segnet, etc.
  ├── result/
  │   └── quantitative & qualitative result
  ├── utils/
  │   ├── datasets.py
  │   ├── metrics.py
  │   ├── preprocess.py
  │   ├── runner.py
  │   └── vision.py
  │  
  ├── FCNs.py
  ├── FPN.py
  ├── UNet.py
  ├── MC-FCN.py
  ├── BR-Net.py
  ├── ResUNet.py
...

Files for generate visualization

├── visSingle.py
├── visSingleComparison.py
...

Model Performance

Performance

  • Overall accuracy, precision, recall, f1-score, jaccard index(IoU) and kappa coefficient performance

Computational efficiency

  • Train and Test FPS time

Visualization Samples

Learning Curve

  • FCN8s FCN8s training curve

Segmentation and outline extraction

  • FCN8s FCN8s segmentation maps

Segmentation result comparison

  • FCN32s, FCN16s, FCN8s FCN8s, FCN16s, FCN32s

Edge extraction result comparison

  • FCN32s, FCN16s, FCN8s FCN8s, FCN16s, FCN32s

Citation

If it helps, please cite the paper.

@article{wu2018geoseg,
  title={Geoseg: A Computer Vision Package for Automatic Building Segmentation and Outline Extraction},
  author={Wu, Guangming and Guo, Zhiling},
  journal={arXiv preprint arXiv:1809.03175},
  year={2018}
}