Home

Awesome

<img src="https://github.com/BarCodeReader/RCSB-PyTorch/blob/main/bin/RCSB_logo.png" alt="drawing" width="1000"/>

PWC PWC PWC PWC PWC PWC

[WACV-2022]Recursive Contour Saliency Blending Network for Accurate Salient Object Detection

News

Our latest Transformer based SOTA: SelfReformer

Network Architecture

network

Prerequisites

Ubuntu 18.04
Python==3.8.3
Torch==1.8.0+cu111
Torchvision=0.9.0+cu111
Kornia

Dataset

For all datasets, they should be organized in below's fashion:

|__dataset_name
   |__Images: xxx.jpg ... 
   |__Masks : xxx.png ...

For training, put your dataset folder under:

dataset/

For evaluation, download below datasets and place them under:

dataset/benchmark/

Suppose we use DUTS-TR for training, the overall folder structure should be:

|__dataset
   |__DUTS-TR
      |__Images: xxx.jpg ... 
      |__Masks : xxx.png ... 
   |__benchmark
      |__ECSSD
         |__Images: xxx.jpg ... 
         |__Masks : xxx.png ... 
      |__HKU-IS
         |__Images: xxx.jpg ... 
         |__Masks : xxx.png ... 
      ...

Train & Test

Firstly, make sure you have enough GPU RAM.
With default setting (batchsize=4), 24GB RAM is required, but you can always reduce the batchsize to fit your hardware.

Default values in option.py are already set to the same configuration as our paper, so
to train the model, simply:

python main.py --GPU_ID 0

to test the model, simply:

python main.py --test_only --pretrain "bal_bla.pt" --GPU_ID 0

If you want to train/test with different settings, please refer to option.py for more control options.
Currently only support training on single GPU.

Pretrain Model & Pre-calculated Saliency Map

Our pretrain model and pre-calculated saliency map: [Google]

If you have problem loading the model due to latest torch use zip file as serialization, download the "RCSB_old_style.pt" instead. It is the same as "RCSB.pt", just to fit older torch versions.

Evaluation

Firstly, obtain predictions via

python main.py --test_only --pretrain "bal_bla.pt" --GPU_ID 0 --save_result

Output will be saved in ./output/ by default.

For PR curve and F curve, we use the code provided by this repo: [BASNet, CVPR-2019]
For MAE, F measure, E score and S score, we use the code provided by this repo: [F3Net, AAAI-2020]

Evaluation Results

Qualitative Results

pred contour

Quantitative Results

mae_table prfm_curve

Citation

If you like this work, please cite our paper

@InProceedings{Ke_2022_WACV,
    author    = {Yun, Yi Ke and Tsubono, Takahiro},
    title     = {Recursive Contour-Saliency Blending Network for Accurate Salient Object Detection},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2022},
    pages     = {2940-2950}
}

Contribution

If you want to contribute or make the code better, simply submit a Pull-Request to develop branch.