Skip to content

Training option #13

@byeongjuhan

Description

@byeongjuhan

Hi, authors.

I read your reflection removal paper interestingly and trained your method newly by using your code and the default training option. However, the trained model does not show the performance of the shared pre-trained model.

For training, I prepare the training dataset containing `trasmission_layer' and 'reflection_layer' images in Zhang et al.'s synthetic dataset and 200 real images in Nature and 90 (actually 89) images in Zhang et al.'s real dataset. I think the training set was made as described in the paper. I now doubt the training option caused different results.

May I ask you to provide the hyper-parameter setting used in the paper?
(Or can you tell me any other causes of the worse performance? )

Thanks.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions