This package contains the source code for A Bi-directional Mssage Passing Model for Salient Object Detection, CVPR 2018. This code is tested on Tensorflow 1.2.1 Ubuntu14.04.

Test

• Instill these requirements if necessary: Python 2.7, Tensorflow 1.2.1, Numpy, Opencv.
• Put your test images in the ./data directory.
• Download the pretrained model from here, and put it under the ./model directory.
• Run TestingModel.py to generate saliency map.

Train

• Built a train_list.txt for your training data, and revise the data path in TrainingModel.py.
• Run TrainingModel.py for training the saliency model.

Saliency map of this paper can be downloaded BaiduYun, GoogleDrive.

@InProceedings{Zhang_2018_CVPR,
    author = {Zhang, Lu and Dai, Ju and Lu, Huchuan and He, You and Wang, Gang},
    title = {A Bi-Directional Message Passing Model for Salient Object Detection},
    booktitle = CVPR,
    year = {2018}}

Steps:

• Download vgg16.npy from https://mega.nz/#!YU1FWJrA!O1ywiCS2IiOlUCtCpI6HTJOMrneN-Qdv3ywQP5poecM, and put to ./
• Donwload pretrained model from https://pan.baidu.com/s/1ZSUW8YPvLR9mRjZ7_ISVnw, and put to ./model
• python TestingModel.py