[FFmpeg-devel] [FFmpeg-cvslog] Adds ESPCN super resolution filter merged with SRCNN filter.

Pedro Arthur bygrandao at gmail.com
Tue Jul 3 19:26:41 EEST 2018


2018-07-03 12:25 GMT-03:00 Daniel Oberhoff <danieloberhoff at googlemail.com>:
> If I may pitch in: why not separate issues? I could imagine a generic convolutional nn filter which gets given topology and weights via some config file. Then provide files separately for this particular trained network.
That's more or less what we already have, we have generic cnn
machinery and we are already able to load models(weights) from a file.
But we have also default weights, in case the user does not have/want
to use an external file.
It is reasonable we provide a default builtin model, I think it is
unreasonable to mandatory require the user to have an external file
with weights to just upscale an image with the sr filter.

Anyway we are working to address the concerns pointed out, soon I'll
be sharing a repo with all the scripts needed to perform training and
model generation.


More information about the ffmpeg-devel mailing list