ganwang / visalientobject

Video Salient Object Detection via Fully Convolutional Networks (TIP18)

Keras Implementation (v1) for

Video salient object detection via fully convolutional networks, IEEE Trans. on Image Processing, 27(1):38-49, 2018

By Wenguan Wang and Jianbing Shen and Ling Shao

========================================================================

Please install keras first

The model weights and results on DAVIS and FBMS datasets can be downloaded from

Baidu Wangpan:https://pan.baidu.com/s/1dE22MFR

password: jmph

or from google drive: https://drive.google.com/drive/u/0/folders/1mc6nnr8RrMZwAXjV0XS9Hv7QJE4quvkx

Please put the models under the folder 'videosalientobjectdetection' and run main.py.

Our results (MAE) are sightly improved over the original scores reported in our paper:

DAVIS: 0.0587 FBMS: 0.0619

You can find our results on UVSD dataset here: https://pan.baidu.com/s/1DUWHiIf-EGG1l-izaAf09Q

Our code for synthesizing video data from still image has been uploaded.

UnZip synthesizevideo.rar first.

Then install vlfeat toolbox first (in 'functions' folder).

Run synthesizevideo.m.

========================================================================

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

1. W. Wang, J. Shen, and L. Shao, Video salient object detection via fully convolutional networks, IEEE Trans. on Image Processing, 27(1):38-49, 2018

2. W. Wang, J. Shen, and L. Shao, Consistent video saliency using local gradient flow optimization and global refinement, IEEE Trans. on Image Processing, 24(11):4185-4196, 2015

3. W. Wang, J. Shen, R. Yang, and F. Porikli, Saliency-aware video object segmentation, IEEE Trans. on Pattern Analysis and Machine Intelligence, 40(1):20-33, 2018

4. W. Wang, J. Shen, F. Porikli, Saliency-aware geodesic video object segmentation, IEEE CVPR, pp. 3395-3402, 2015

========================================================================

Any comments, please email: [email protected]