Skip to content
Deep Visual Attention Prediction (TIP18)
Branch: master
Clone or download
Latest commit e2e20fa Mar 21, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
datasets/pascal Add files via upload Dec 17, 2017
models Add files via upload Dec 26, 2017
attention_config.m Add files via upload Dec 17, 2017
deep_attention_test.m Add files via upload Dec 17, 2017
startup.m Add files via upload Dec 17, 2017

This repo contains a CAFFE re-implementation for

"Deep Visual Attention Prediction", TIP18

By Wenguan Wang, Jianbing Shen


Our current implementation is based on the caffe version of Holistically-Nested Edge Detection (has been included in this repository, see external/caffe). We also use some functions from fasterrcnn's caffe. But it can be easily implemented in standard caffe library.

We plan to re-implement it on Keras, but it depends on my schedule.

a. Please first compile the 'Caffe' for matlab.

b. Please download our MODEL and RESUTLS on DUT, MIT300, MIT1003, Pascal, and Toronto datasets from

google drive:

or Baidu Wangpan: password: wdwa

and put the model 'attention_final' into 'models' folder.

c. Run 'deep_attention_test' and the results can be found in 'datasets/results/'.

If you find our method useful in your research, please consider citing:

    Author = {Wenguan Wang, Jianbing Shen},
    Title = {Deep Visual Attention Prediction},
    Journal = {IEEE Transactions on Image Processing},
    Year = {2018}


Any comments, please email:

You can’t perform that action at this time.