Joint Salient Object Detection and Camouflaged Object Detection via Uncertainty-aware Learning

This code is the journal extension version of: Uncertainty-aware Joint Salient Object and Camouflaged Object Detection (CVPR2021)

Train the model

Set up

pip install -r requirements.txt  

Train Model

• Prepare data for training (We provided the related data in:[Google_Drive]. Please download it and put it in the './train_data/' folder)
  • 'duts+wmae' means the augmented SOD training set
  • 'COD_train' means the COD training dataset
  • 'JPEGImages_select' means the selected PASCAL VOC dataset

python train.py  

Test and Evaluate Model

• Prepare data for testing (We provided the related SOD and COD test image and ground-truth in:[Google_Drive]. Please download it and put it in the './test_data/' folder)

python test.py   

Pretrained model and Prediction Maps

Trained model:

Please download the trained model and put it in "./models/": [Google_Drive];

Prediction Maps:

Results of our model on four benchmark datasets (CAMO, CHAMELEON, COD10K, NC4K) for COD:[Google_Drive]; and six benchmark datasets (DUTS, ECSSD, DUT, HKU-IS, PASCAL, SOD ) for SOD: [Google_Drive]

Thanks Yuxin Mao for the Evaluation code link