Which model should I use to test on the waterloo dataset? about dbcnn

[zwx8981]

@Xiaocong6 Hi, the folder "models" contains training-testing split files for each database. The scnn.mat under the 'data/models' folder is trained using the combination of Waterloo and PASCAL as we mentioned in the paper. The model used for Waterloo experiments is not released in this repo.

[Xiaocong6]

Thanks!
I am not familiar with Matlab. I want to ask, if I want to use the models your provided to test the results on my own dataset, which file should I run?

[zwx8981]

@Xiaocong6 You can train a model on a dataset by running run_exp.m first, then you can use the trained model to test on your dataset. See the demo.m for a simple demo of predicting the quality score of an input image.

[Xiaocong6]

Thanks!
Whick dataset is used to train the model used for Waterloo experiments?

[zwx8981]

@Xiaocong6 For fair comparison, we train the model on the LIVE dataset for Waterloo experiments.

[Xiaocong6]

Thank you very much! I got it.