window10 or Ubuntu,
cuda10.1 + cudnn 7.6.5 + python3.6,
torch1.7.1 + torchvision 0.8.2,
matplotlib, sklearn, scikit-image, opencv-python
1.you need to find the file "dataset.py" in lines 72,73,74
and modify the path to the storage path for your project
line72 self.train_root = "/your_path/data/oil-spill-detection-dataset/train"
line73 self.val_root = "/your_path/data/oil-spill-detection-dataset/val"
line74 self.test_root = "/your_path/data/oil-spill-detection-dataset/test"
2.you need to see the lines 346,347,348 in "main.py"
If you use linux system to run this project, you can use line346 and line347.
If you use windows system to run this project, you can use line346 and line348.
and then, you can get the predict results.
line346 predict =Image.fromarray(predict).convert('L') #predict
line347 #predict.save(dir +'/'+mask_path[0].split('/')[-1]) #linux
line348 predict.save(dir +'\\'+mask_path[0].split('\\')[-1]) #win
3.you need to see the lines 350,351,352 in "main.py"
If you use linux system to run this project, you can use line350 and line351.
If you use windows system to run this project, you can use line350 and line352.
If you want to get the differences between the predict image and the ground truth,
you can delete # in line350,351 or line350,352 .
line350 #i_u_img = Image.fromarray(i_u_img).convert('L') #difference
line351 #i_u_img.save(dir +'/'+mask_path[0].split('/')[-1]) #linux
line352 #i_u_img.save(dir +'\\'+mask_path[0].split('\\')[-1]) #win
4.you need to download the weight files what you need from this website.
And you need to put the weight files to the path " ./saved_model/oil_spill_detection/"
Baidu netdisk: https://pan.baidu.com/s/15N5OctWqZQ7Js7Sg37f9zA
Extraction code: kdy4
5.You can run this project by using the command line, just like this.
python main.py --action test --arch Attention_UNet --epoch 51 --batch_size 4
in --arch, you can use UNet, UNet_AGC, unet++, unet++_AGC, Attention_UNet, Attention_UNet_AGC
r2unet, r2unet_AGC, fcn8s, fcn8s_AGC, fcn32s, fcn32s_AGC
UNet_AGC means the UNet with Gamma_Log Net.
you will get the results in "saved_predict" folder. The predict results and the difference images are in the same place.
The specific values of iou and difference are displayed at the interface.