[FFmpeg-devel] [PATCH 1/7] libavfilter: Adds on the fly generation of default DNN models for tensorflow backend instead of storing binary model.

[Pedro Arthur]

2018-08-07 13:34 GMT-03:00 Rostislav Pehlivanov <atomnuker at gmail.com>:
> On 7 August 2018 at 17:32, Pedro Arthur <bygrandao at gmail.com> wrote:
>
>> 2018-08-06 18:09 GMT-03:00 Sergey Lavrushkin <dualfal at gmail.com>:
>> > Changed sprintf to snprintf.
>> Patch pushed.
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel at ffmpeg.org
>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>
> Why are the weights changed? Sure, some have been removed, in fact a lot,
The data is the same, I think just the formatting (spaces) changed.

> but at the end we agreed that there should be absolutely no binary weights
> left in lavf, and everything should be loadable on runtime.
> Is this going to be implemented with a later patchset?
We are removing things gradually, I'm still thinking in how to
distribute the filter data.
Do you have any tips on where it would be easily accessible for users?

> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel