MSCNet: A lightweight multi-scale context network for Salient Object Detection in Optical Remote Sensing Images

Code for ICPR 2022 paper "A lightweight multi-scale context network for Salient Object Detection in Optical Remote Sensing Images", by Yuhan Lin, Han Sun, Ningzhong Liu, Yetong Bian, Jun Cen, and Huiyu Zhou

Requirement

python-3.6
pytorch-1.8.1
torchvision
numpy
tqdm
cv2

Usage

• Clone this repo into your workstation git clone https://github.com/NuaaYH/MSCNet.git
• The datasets used in this paper can be download from BaiduYun: https://pan.baidu.com/s/1oDOsPZkRisCBzAtoY6SDMA (code:hl0s)
• Set the project format as follows:
  ./MSCNet
  ./Dataset
• Create the folders in MSCNet as shown below:
  ./Outputs/pred/MSCNet/EORSSD(or ORSSD)/Test
  ./Checkpoints/trained

training

1. Comment out line 57 of run.py like #self.net.load_state_dict......
2. Comment out line 173 of run.py like #run.test() and ensure that the run.train() statement is executable
3. python run.py

testing

1. Put the model weights in ./Checkpoints/trained and ensure that line 57 of run.py is executable
2. Comment out line 172 of run.py like #run.train() and ensure that the run.test() statement is executable
3. python run.py

evaluation

The evaluation code can be available at https://github.com/zyjwuyan/SOD_Evaluation_Metrics.

Results

• The results of ours and the comparison methods in our paper can be download from BaiduYun:
  https://pan.baidu.com/s/1Vl7maJjQiQVS8VG77t5lLw (code:arjc)
• The pre-trained model can be download from BaiduYun:
  https://pan.baidu.com/s/1AcQJjKdsk3SWPr2O6i-c0A (code:fkci)