[FFmpeg-devel] [PATCH] libavfilter: Add derain filter init version--GSoC Qualification Task.

[Pedro Arthur]

Em qui, 11 de abr de 2019 às 02:55, <xwmeng at pku.edu.cn> escreveu:
>
> We made some modifications on the original ESPCN model, such as change the input image from one channel(Y) to three channels(RGB) and remove the up-sampling procedure. The model file has been uploaded in https://github.com/XueweiMeng/derain_filter and you can download the training/testing dataset from http://www.icst.pku.edu.cn/struct/Projects/joint_rain_removal.html. I didn't save the PSNR/SSIM score during the training and evaluating process. So the data will be uploaded later.
>

Indeed the model is not the espcn anymore, as it would imply the use
of the up-sampling layer. I think it is better to label this network
as a generic convolutional network and just describe its layout
(number of layers and layer dimension). Using the espcn name is
misleading.

Please, always include your training results and relevant observations
when sending a patch otherwise it is hard to evaluate your work if we
did not even know what to expect from the output.