FactSeg: Foreground Activation Driven Small Object Semantic Segmentation in Large-Scale Remote Sensing Imagery

by Ailong Ma, Junjue Wang*, Yanfei Zhong* and Zhuo Zheng

This is an official implementation of FactSeg in our TGRS paper " FactSeg: Foreground Activation Driven Small Object Semantic Segmentation in Large-Scale Remote Sensing Imagery"

Citation

If you use FactSeg in your research, please cite our coming TGRS paper.

@ARTICLE{FactSeg,
  author={Ma, Ailong and Wang, Junjue and Zhong, Yanfei and Zheng, Zhuo},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={FactSeg: Foreground Activation-Driven Small Object Semantic Segmentation in Large-Scale Remote Sensing Imagery}, 
  year={2022},
  volume={60},
  number={},
  pages={1-16},
  doi={10.1109/TGRS.2021.3097148}}

This is follow-up work of our FarSeg (CVPR2020).

@inproceedings{zheng2020foreground,
  title={Foreground-Aware Relation Network for Geospatial Object Segmentation in High Spatial Resolution Remote Sensing Imagery},
  author={Zheng, Zhuo and Zhong, Yanfei and Wang, Junjue and Ma, Ailong},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={4096--4105},
  year={2020}
}

Getting Started

Install SimpleCV

pip install --upgrade git+https://github.com/Z-Zheng/SimpleCV.git

Requirements:

• pytorch >= 1.1.0
• python >=3.6

Prepare iSAID Dataset

ln -s </path/to/iSAID> ./isaid_segm

Evaluate Model

1. download pretrained weight in Google Drive

2. move weight file to log directory

mkdir -vp ./log/
mv ./factseg50.pth ./log/model-60000.pth

3. inference on iSAID val

bash ./scripts/eval_factseg.sh

Train Model

bash ./scripts/train_factseg.sh