Boundary-Semantic Collaborative Guidance Network with dual-stream feedback mechanism for Salient Object Detection in Optical Remote Sensing Imagery

Network Architecture

Accuracy

Saliency maps

We provide saliency maps of our BSCGNet on ORSSD, EORSSD and ORSI-4199 in './saliency_maps/Ours.7z'.

Training

Backbone: VGG16 (code: 7a1s)

Our BSCGNet can be found in './models/BSCGNet/Mynet.py'

Modify paths of datasets, then run train_net.py.

Testing

modify paths of pre-trained model and datasets, then run test_net.py.

Evaluation Tool

run evaluation.py to evaluate the above saliency maps.

Related Works

we provide a common code base which has 17 SOD methods, including 8 ORSI-SOD methods, 7 NSI-SOD methods, 2 traditional SOD methods and the method proposed in this
paper.You can find them directly in './models/'.

The references are as follows.

• Traditional methods

DSG: (TIP 2017) Salient region detection using diffusion process on a two-layer sparse graph

RCNN: (TIP 2018) Reversion Correction and Regularized Random Walk Ranking for Saliency Detection

• DL-based NSI-SOD

SCRN: (ICCV 2019) Stacked cross refinement network for edge-aware salient object detection

CPDNet: (CVPR 2019) Cascaded partial decoder for fast and accurate salient object detection

GateNet: (ECCV 2020) Suppress and balance: A simple gated network for salient object detection

CTDNet: (ACMM 2021) Complementary trilateral decoder for fast and accurate salient object detection

PFSNet: (AAAI 2021) Pyramidal feature shrinking for salient object detection

BSANet: (AAAI 2022) I can find you! Boundary-guided Separated Attention Network for Camouflaged Object Detection

PGNet: (CVPR 2022) Pyramid Grafting Network for One-Stage High Resolution Saliency Detection

• DL-based ORSI-SOD

LVNet: (TGRS 2019) Nested network with two-stream pyramid for salient object detection in optical remote sensing images.

FSMINet: (GRSL 2022) Fully Squeezed Multiscale Inference Network for Fast and Accurate Saliency Detection in Optical Remote-Sensing Images

ERPNet: (TCYB 2022) Edge-guided recurrent positioning network for salient object detection in optical remote sensing images

CorrNet: (TGRS 2022) Lightweight salient object detection in optical remote sensing images via feature correlation

SeaNet: (TGRS 2023) Lightweight Salient Object Detection in Optical Remote-Sensing Images via Semantic Matching and Edge Alignment

SRAL: (TGRS 2023) Distilling Knowledge From Super-Resolution for Efficient Remote Sensing Salient Object Detection

EMFINet: (TGRS 2022) Edge-aware multiscale feature integration network for salient object detection in optical remote sensing images

JRBM: (TGRS 2022) ORSI salient object detection via multiscale joint region and boundary model

Citation

If these codes are helpful for you, please cite this paper:

@ARTICLE{10315168,
  author={Feng, Dejun and Chen, Hongyu and Liu, Suning and Liao, Ziyang and Shen, Xingyu and Xie, Yakun and Zhu, Jun},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Boundary-semantic collaborative guidance network with dual-stream feedback mechanism for salient object detection in optical remote sensing imagery}, 
  year={2023},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TGRS.2023.3332282}}

Contact

If you encounter any problems with the code.

Please contact me at chy0519@my.swjtu.edu.cn.