zijundeng / r3net

Code for the IJCAI 2018 paper "R^3Net: Recurrent Residual Refinement Network for Saliency Detection"

R³Net: Recurrent Residual Refinement Network for Saliency Detection

by Zijun Deng, Xiaowei Hu, Lei Zhu, Xuemiao Xu, Jing Qin, Guoqiang Han, and Pheng-Ann Heng [paper link]

This implementation is written by Zijun Deng at the South China University of Technology.

---

Citation

@inproceedings{deng18r,
     author = {Deng, Zijun and Hu, Xiaowei and Zhu, Lei and Xu, Xuemiao and Qin, Jing and Han, Guoqiang and Heng, Pheng-Ann},
     title = {R$^{3}${N}et: Recurrent Residual Refinement Network for Saliency Detection},
     booktitle = {IJCAI},
     year = {2018}
}

Saliency Map

The results of salienct object detection on five datasets (ECSSD, HKU-IS, PASCAL-S, SOD, DUT-OMRON) can be found at Google Drive.

Trained Model

You can download the trained model which is reported in our paper at Google Drive.

Requirement

• Python 2.7
• PyTorch 0.4.0
• torchvision
• numpy
• Cython
• pydensecrf (here to install)

Training

1. Set the path of pretrained ResNeXt model in resnext/config.py
2. Set the path of MSRA10K dataset in config.py
3. Run by python train.py

The pretrained ResNeXt model is ported from the official torch version, using the convertor provided by clcarwin. You can directly download the pretrained model ported by me.

Hyper-parameters of training were gathered at the beginning of train.py and you can conveniently change them as you need.

Training a model on a single GTX 1080Ti GPU takes about 70 minutes.

Testing

1. Set the path of five benchmark datasets in config.py
2. Put the trained model in ckpt/R3Net
3. Run by python infer.py

Settings of testing were gathered at the beginning of infer.py and you can conveniently change them as you need.

Useful links

• MSRA10K: our training set
• ECSSD, HKU-IS, PASCAL-S, SOD, DUT-OMRON: the five benchmark datasets

r3net's Issues

你好，我看你训练的loss函数是BCEWithLogitsLoss，为什么还要在model里最后输出里加上sigmoid函数

The model file

Can you tell me why the model can't be modified?
For example, After i changed the stride in resnet_101_32×4d.py, it would be wrong since the change

Optimizer selection

Hello, first of all, thank you very much for your code.May I ask if your optimizer has tried Adam and RMSprop?I think the optimizer you used in your code is SGD. May I ask whether it was selected after testing, or just this one?I feel in the training process using SGD in the training speed is relatively slow, but the effect is quite good.

参考文献样式

您好，请问能提供一份IJCAI参考文献的endnotes样式给我吗？

How to use the trained model to get the map of the datasets

Novel Methods

Congralation! However, please make a big contribution for our papers in the future. The existing works in your group seem that they are mostly modified from previous pubilic works in other group! This is not a good idea for community.

No module 'resnext101

Hello, thank you for sharing the code for us to learn the paper.When I run the program,there is a probelm ,which is"File "D:three\R3Net-master\resnext_init_.py.,line 1,in from resnext101 import ResNext101, ModuleNotFoundError:No module 'resnext101'. ".resnext101.py exists. I don't understand why this is a mistake,.May I ask you for help?thank you very much.

请问能否提供一下DUT_TEST数据集的saliency map吗？

请问能否提供一下DUT_TEST数据集的saliency map吗？
就是R3Net的对这个数据集的测试结果

I test the give 6000.pth model on the ECSSD dataset, Fmeasure are 0.7323. I trained myself get the results 0.933. Can you check the 6000.pth is the right model?

I test the give 6000.pth model on the ECSSD dataset, Fmeasure are 0.7323. I trained myself get the results 0.933. Can you check the 6000.pth is the right model?

Inference: ZeroDivisionError: float division by zero

Are there any fixes for this?

~/R3Net$ python infer.py load snapshot '6000' for testing rwh ./ckpt R3Net 6000.pth Traceback (most recent call last): File "infer.py", line 102, in <module> main() File "infer.py", line 93, in main [rrecord.avg for rrecord in recall_record]) File "/home/ubuntu/R3Net/misc.py", line 67, in cal_fmeasure max_fmeasure = max([(1 + beta_square) * p * r / (beta_square * p + r) for p, r in zip(precision, recall)]) ZeroDivisionError: float division by zero