Awesome

PDNet

For RGBD salient object detection, the paper has benn accepted at icme2018 and the paper link is here

The pre_trained model

Baidu Yun: https://pan.baidu.com/s/1rGXSTxtU8iNQa7FPGP83Wg password: 9hbq

with deep channel: https://pan.baidu.com/s/1Uu03Z_1R6_9BpSZfOzK31g password: cpe8

modify the test.py file.(I am sorry for forgetting that)

Google drive: https://drive.google.com/open?id=18yXaBAeG_6qFuk4wyViSWiEXog5ctabM

The pre-trained model based on MSRA10k and DUTS-TR datasets, which contain about 20k pictures.

Train

Download the dataset, for example, MSRA10k, and run the preprocess script

python preprocess.py

then run

python train.py

Test

Download the pre-trained model and the test picture, run

python test.py

The results

If you use this code. Please cite as:

PDNet: Prior-Model Guided Depth-enhanced Network for Salient Object Detection. Chunbiao Zhu, Xing Cai, Kan Huang, Thomas.H Li and Ge Li IEEE International Conference on Multimedia and Expo (ICME) 2018. or

@INPROCEEDINGS{PDNet2018,
author=
"Zhu, Chunbiao
and Cai, Xing
and Kan, Huang
and Thomas.H, Li
and Ge, Li",
booktitle={2018 International Conference on Multimedia and Expo},
title={PDNet: Prior-model Guided Depth-enhanced Network for Salient Object Detection},
year={2019},
}