sr
sr copied to clipboard
Ffmpeg ready, trained models.
I'm attaching the pretrained model/header files here. This should be helpful for the people that don't have the resources to train the models themselves. (Takes hours even on a dgx server)
Models are trained using the default parameters defined in their respective train_XXX.sh
files.
In addition to that, there's a second version included for the VESPCN model, the difference is vespcn-mc
uses --use_mc
flag for motion compensation in addition to the default parameters. (only model that can use that flag)
However, as of now. I can only get ffmpeg to use SRCNN and ESPCN filters, as others are apparently not supported yet
$ sha256sum dnn_models.tar.gz
ac3e1b20bb942a3156042a07b7e68ed2aec66f49d92b48f5a4dfbf6cb5283417 dnn_models.tar.gz