RGB-T-Salient-Object-Detection-via-CNN-Features-and-Result-Saliency-Maps-Fusion

This project provides the codes and results for 'RGB-T-Salient-Object-Detection-via-CNN-Features-and-Result-Saliency-Maps-Fusion.'

Overview

Code

For the training of the proposed network:

download pretrained vgg model from link,code：0000; put it in model directory.

change img_root in train.py to load train data .

use python train.py to start training.

For the testing of the proposed network:

change model_path and root in test.py to load trained checkpoint and test data.

change out_path in test.py to save test results.

use python test.py to start testing.

For Result Fusion:

change paths in resultFusion.py.

use python resultFusion.py to fuse saliency maps from RGB and T modalities, there are some examples in asserts for testing.

Experimental Results

Result of VT5000 dataset: link, code：1234.

Result of VT1000 dataset: link, code：4321.

The evaluation toolbox is provided by https://github.com/ArcherFMY/sal_eval_toolbox