SdBAN: Salient Object Detection Using Bilateral Attention Network with Dice Coefficient Loss (IEEE ACCESS) [Link]

Tensorflow based keras implementation of "SdBAN: Salient Object Detection Using Bilateral Attention Network with Dice Coefficient Loss"

Qualitative Evaluation

Quantative Eavaluation

Getting Started

Installation

• Clone thos repository

git clone https://github.com/tiruss/Salient_Code.git

• You can install all the dependencies by

pip install -r requirements.txt

Download datasets

• Download training datasets [DUTS-TR] from the link

• Download [HKU-IS] for test from the link

• Other datasets can download from the link [sal_eval_toolbox] Thank you for the awesome evaluation toolbox!

Run experiments from pretrained weight

• Download pretrained weight from the link

• [Google drive] [Baidu drive] Baidu drive will be updated soon.

• Run test.py

python test.py --weight [pretrained weight] --input_dir [test_img_dir] --output_folder "outputs"

• Pre-computed salinecy maps can download from the link

• [Google drive] [Baidu drive] Baidu drive will be updated soon.

Train from scratch

• DUTS-TR is our traning set for pair comparison

• Run train.py

python train.py --img_folder [DUTS-TR img dir] --label_folder [DUTS-TR label dir] --epoch --batch_size --num_gpu

Citation

@ARTICLE{9107080,
  author={D. {Kang} and S. {Park} and J. {Paik}},
  journal={IEEE Access}, 
  title={SdBAN: Salient Object Detection Using Bilateral Attention Network With Dice Coefficient Loss}, 
  year={2020},
  volume={8},
  number={},
  pages={104357-104370},}