[mohit-0212]

I still think we should distribute the machine learning model that we've trained for a show.

Yes, what we can do is bundle our models together separately as sort of an api, and then use the input from the user (video frames/audio) for querying to it. 

but I don't know if there is something similar for audio at the moment.

So if there isn't anything for the audio, would we have to write a cpp file for it (similar to what we have as rendercapture.cpp)? If yes, how should we go about it? Or is there any workaround for it.

[Razze]

Well we might have a problem here, as the renderCapture Api is likely to get deprecatd due to the bad design/shape it is in.
@FernetMenta Might know more about it, as he brought it up in our gsoc slack channel.

Not really an idea on how to go on now, as directly accessing the users video files seems "weird" at least and might be something that would keep an addon out of our repo. @Martijn for reference.

[mohit-0212]

Hey

Any updates/suggestions on how can we proceed? [ Can we somehow take the input using the microphones if we use the audio signatures? Would that be viable? ]

Also I have started to work on my proposal, but we need to figure out this proper deployment issue for that also.

[Razze]

No, theres no microphone support in kodi.
This pretty much feels like a dead end, unless someone starts to implement something in core that can replace renderCapture.

[mohit-0212]

So how do you suggest I proceed with my proposal?
I had started working on a rough implementation of the audio fingerprinting algorithm taking help from the paper, http://www.ee.columbia.edu/~dpwe/papers/...shazam.pdf

Should I look in the existing renderCapture files and see what can we do about it?

[Razze]

You will need to speak with @FernetMenta on how to proceed with replacing renderCapture. I would suggest the #kodi-gsoc channel for that.

[mohit-0212]

Hey "Razze" 
I am working on a proposal based on the discussions on the irc channel and this forum post here. I will try to submit it within 1-2 days, I hope it won't be late.
Should I post my proposal here as well for review apart from the gsoc portal as I could see it in few previous year posts?

[Razze]

Well it would probably be good for transparency if you post it here too. But I will only review on the gsoc portal for now.

[mohit-0212]

Hey
I have uploaded my draft on the GSoC portal for your feedback and inputs.

Thanks

[docwra]

Someone once asked me to write a metadata site to store all the intro/outtro timings.

I didn't think much more about it at the time as I had other priorities, but thought it was worth mentioning in this discussion.

[mohit-0212]

Hey

Thanks for your input. But what I think is that each user might have a different type of video file for the same episode with slightly different durations, so a general metadata site, in my opinion, won't work well.
Correct me, if I misunderstood what you were trying to do.

Thanks

[docwra]

Hey

Thanks for your input. But what I think is that each user might have a different type of video file for the same episode with slightly different durations, so a general metadata site, in my opinion, won't work well.
Correct me, if I misunderstood what you were trying to do.

Thanks

Yes that's certainly an issue but nothing that can't be fixed with scraping, file hashing or simple versioning. You would actually be surprized how little versions there are of TV shows floating around the internet.

EDIT: I personally rejected the idea for similar reasons, but thought it was worth mentioning.

[Razze]

Hey

Thanks for your input. But what I think is that each user might have a different type of video file for the same episode with slightly different durations, so a general metadata site, in my opinion, won't work well.
Correct me, if I misunderstood what you were trying to do.

Thanks

Yes, that's the core problem and the reason why we want to do models and deploy those via REST or similar. But we need to find someone to host those or do it on our own.

[mohit-0212]

Hey @Razze
How does the overall proposal look like? Should I add or modify anything else, apart from the few things you are reviewing?

[Razze]

I would have said something if I would think there was something to add