Accepted paper in IEEE Trans on Cybernetics, 'PANet: Patch-Aware Network for Light Field Salient Object Detection', Yongri Piao, Yongyao Jiang, Miao Zhang, Jian Wang and Huchuan Lu.

Windows 10

PyTorch 1.4.0

CUDA 10.0

Cudnn 7.6.0

Python 3.6.5

Numpy 1.16.4

Modify your path of training dataset in config.py

Run train.py for training the saliency model, the maximum of training iterations is 500000.

Run train_mslm.py for training the MSLM model, the maximum of training iterations is 5000.

Run train_srm.py for training the SRM model, the maximum of training iterations is 5000.

Run train_second_decoder.py for training the second decoder, the maximum of training iterations is 500000.

Download pretrained models from here. Code: qwer

Modify your path of testing dataset in config.py

Run test to inference saliency maps

DUTS-LFSD&HFUT-LFSD&LFSD, Download link. Code: qwer

Contact:Yongyao Jiang. Email:[email protected]