conet's Introduction

CoNet

Code repository for our paper entilted "Accurate RGB-D Salient Object Detection via Collaborative Learning" accepted at ECCV 2020 (poster).

Overall

CoNet Code

> Requirment

pytorch 1.0.0+
torchvision
PIL
numpy

> Usage

1. Clone the repo

git clone https://github.com/jiwei0921/CoNet.git
cd CoNet/

2. Train/Test

test
Our test datasets link and checkpoint link code is 12yn. You need to set dataset path and checkpoint name correctly.

'--phase' as test in demo.py
'--param' as True in demo.py

python demo.py

train
Our training dataset link code is 203g. You need to set dataset path and checkpoint name correctly.

'--phase' as train in demo.py
'--param' as True or False in demo.py
Note: True means loading checkpoint and False means no loading checkpoint.

python demo.py

> Results

We provide saliency maps (code: qrs2) of our CoNet on 8 datasets (DUT-RGBD, STEREO, NJUD, LFSD, RGBD135, NLPR, SSD, SIP) as well as 2 extended datasets (NJU2k and STERE1000) refer to CPFP_CVPR19.

Note: For evaluation, all results are implemented on this ready-to-use toolbox.

> Related RGB-D Saliency Datasets

All common RGB-D Saliency Datasets we collected are shared in ready-to-use manner.

The web link is here.

If you think this work is helpful, please cite

@InProceedings{Wei_2020_ECCV,       
   author={Ji, Wei and Li, Jingjing and Zhang, Miao and Piao, Yongri and Lu, Huchuan},  
   title = {Accurate {RGB-D} Salient Object Detection via Collaborative Learning},     
   booktitle = {European Conference on Computer Vision},     
   year = {2020}     
}

For more info about CoNet, please read the Manuscript.
Thanks for related authors to provide the code or results, particularly, Deng-ping Fan, Hao Chen, Chun-biao Zhu, etc.