[FFmpeg-devel] [PATCH 2/6] avfilter/vf_sr: add options for backend configs

[lance.lmwang at gmail.com]

On Tue, Apr 27, 2021 at 02:27:48AM +0000, Guo, Yejun wrote:
> 
> 
> > -----Original Message-----
> > From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of
> > lance.lmwang at gmail.com
> > Sent: 2021年4月26日 18:49
> > To: ffmpeg-devel at ffmpeg.org
> > Cc: Limin Wang <lance.lmwang at gmail.com>
> > Subject: [FFmpeg-devel] [PATCH 2/6] avfilter/vf_sr: add options for backend
> > configs
> > 
> > From: Limin Wang <lance.lmwang at gmail.com>
> > 
> > Signed-off-by: Limin Wang <lance.lmwang at gmail.com>
> > ---
> >  libavfilter/vf_sr.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/libavfilter/vf_sr.c b/libavfilter/vf_sr.c
> > index 45f941a..282c468 100644
> > --- a/libavfilter/vf_sr.c
> > +++ b/libavfilter/vf_sr.c
> > @@ -55,6 +55,8 @@ static const AVOption sr_options[] = {
> >      { "model", "path to model file specifying network architecture and its
> > parameters", OFFSET(dnnctx.model_filename), AV_OPT_TYPE_STRING,
> > {.str=NULL}, 0, 0, FLAGS },
> >      { "input",       "input name of the model",
> > OFFSET(dnnctx.model_inputname),  AV_OPT_TYPE_STRING,    { .str =
> > "x" },  0, 0, FLAGS },
> >      { "output",      "output name of the model",
> > OFFSET(dnnctx.model_outputname), AV_OPT_TYPE_STRING,    { .str =
> > "y" },  0, 0, FLAGS },
> > +    { "backend_configs",    "backend configs",
> > OFFSET(dnnctx.backend_options),  AV_OPT_TYPE_STRING,    { .str =
> > NULL }, 0, 0, FLAGS },
> > +    { "options",            "backend configs",
> > OFFSET(dnnctx.backend_options),  AV_OPT_TYPE_STRING,    { .str =
> > NULL }, 0, 0, FLAGS },
> >      { NULL }
> >  };
> > 
> 
> vf_dnn_processing.c is designed to cover all the cases for image/video
> processing with M inputs and N outputs (currently only M=1 and N=1 is
> supported), covering the features of vf_sr.c and vf_derain.c. To avoid
> duplicate/similar code in different filters, we don't add more feature in 
> vf_sr.c/vf_derain.c, we just keep supporting their current feature.

The problem is when trying with SRCNN model with vf_sr, I can't find scale_factor
for dnn_processing, so I had to use the old vf_sr still. That's why I had to
make the options be configured for vf_sr like dnn_processing. Or it's easy to
get "out of memory" with default sess_config for TF. 


> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".

-- 
Thanks,
Limin Wang