We provide the resutls of ASIF-Net in our paper. We rename the images, so the name of our result is different from the original data. For your evaluations, we also provide the corresponding renamed GT.
Reults:
Google Drive: https://drive.google.com/open?id=15WlRLFSYG-mQ73DpUngaUqOnUwtz4PPc
Baidu Cloud: https://pan.baidu.com/s/1DVAwqe3n5JeUIuaAkzteYw Password: byxj
GT:
Google Drive: https://drive.google.com/file/d/1UrSt5oc5ER7Ux4Py2rlR58JmCImmEkw9/view?usp=sharing
Baidu Cloud: https://pan.baidu.com/s/12M4COP8C3r1OELul7CtM4Q Password: 1234
We provide our testing code. If you test our model, please download the pretrained model, unzip it, and put the checkpoint to checkpoint/coarse_224 folder.
Pretrained model download:
Google Drive: https://drive.google.com/file/d/1oTYUY9bKEFOrDyXMGTFFrgmFkHGUURM3/view?usp=sharing
Baidu Cloud: https://pan.baidu.com/s/1sQcb2o0of7sQXX3-D8ep5g Password: 1234
TensorFlow implementation of ASIF-Net
Python 3
TensorFlow 1.x
1) normalize the depth maps (note that the foreground should have higher value than the background in our method)
input=(input-min(min(input)))/(max(max(input))-min(min(input)))
The step is very important for accurate results.
2) resize the testing data to the size of 224*224
first normalize depth then resize will be better than first resize depth then normalize in our method. So please strictly follow our steps to generate testing data.
3) put your rgb images to 'test_real' folder and your depth maps to 'depth_real' folder (paired rgb image and depth map should have same name)
python main_test.py
find the results in the 'test_real' folder with the same name as the input image + "_out".
You can use a script to resize the results back to the same size as the original RGB-D image, or just use the results with a size of 224*224 for evaluations. We did not find much differences for the evaluation results.
If you use the results and code, please cite our paper.
@article{ASIF-Net,
title={{ASIF-Net}: Attention steered interweave fusion network for {RGBD} salient object detection},
author={Li, Chongyi and Cong, Runmin and Kwong, Sam and Hou, Junhui and Fu, Huazhu and Zhu, Guopu and Zhang, Dingwen and Huang, Qingming },
journal={IEEE Trans. Cybern.},
volume={PP},
number={99},
pages={1-13},
year={2020}
}
paper link: https://ieeexplore.ieee.org/document/8998588
If you have any questions, please contact Chongyi Li at [email protected].