+- Kodi Community Forum (https://forum.kodi.tv)
+-- Forum: Development (https://forum.kodi.tv/forumdisplay.php?fid=32)
+--- Forum: Kodi Application (https://forum.kodi.tv/forumdisplay.php?fid=93)
+---- Forum: GSoC (https://forum.kodi.tv/forumdisplay.php?fid=299)
+----- Forum: GSOC 2018 (https://forum.kodi.tv/forumdisplay.php?fid=284)
+----- Thread: Interested in GSOC project "Using machine learning to improve suggestions" (/showthread.php?tid=328511)
Pages:
1
2
RE: Interested in GSOC project "Using machine learning to improve suggestions" - Razze - 2018-03-07
Sounds like your on track to me
RE: Interested in GSOC project "Using machine learning to improve suggestions" - amit2rockon - 2018-03-09
@Razze While building this setup on mac i'm getting this error ,
make: the `-j' option requires a positive integral argument/Applications/Xcode.app/Contents/Developer/usr/bin/make -C native/Applications/Xcode.app/Contents/Developer/usr/bin/make -C m4-nativerm -rf x86_64-osx-native/*; mkdir -p x86_64-osx-nativecd x86_64-osx-native; /usr/bin/tar --strip-components=1 -xf /Users/Shared/xbmc-depends/xbmc-tarballs/m4-1.4.18.tar.gztar: (Empty error message)tar: Error exit delayed from previous errors.make[2]: *** [x86_64-osx-native] Error 1make[1]: *** [m4-native] Error 2make: *** [native/.installed-x86_64-osx-native] Error 2
How to get it solved?
RE: Interested in GSOC project "Using machine learning to improve suggestions" - marko.prcac - 2018-03-10
(2018-03-07, 14:15)Razze Wrote: Well first thing would be to setup a local enviroment to debug/compile kodi. You can find infos about that in our WIKI
Hi!
I've tried the windows path but none of the mirrors had all the required packages. I will try ubuntu now, and then return to manual installation of missing packages it this fails too. If you are interested I can provide more info on issues. I am also making a demo app.
RE: Interested in GSOC project "Using machine learning to improve suggestions" - amit2rockon - 2018-03-10
@vedantrathore , did you not face the same problem with building and compiling it
RE: Interested in GSOC project "Using machine learning to improve suggestions" - Razze - 2018-03-11
(2018-03-10, 01:03)marko.prcac Wrote:Well it does happen that you need to run the bat files on windows multiple times. Which isn't good but usually works out.(2018-03-07, 14:15)Razze Wrote: Well first thing would be to setup a local enviroment to debug/compile kodi. You can find infos about that in our WIKI
Hi!
I've tried the windows path but none of the mirrors had all the required packages. I will try ubuntu now, and then return to manual installation of missing packages it this fails too. If you are interested I can provide more info on issues. I am also making a demo app.
RE: Interested in GSOC project "Using machine learning to improve suggestions" - marko.prcac - 2018-03-12
HI @Razze !
I have successfully compiled KODI on Ubuntu and I made my hello world add-on today. I am now on my way with producing a DEMO of my idea as an add-on. Is this something that would be interesting for you to see or do you have any other suggestions for further steps?
BR, Marko
RE: Interested in GSOC project "Using machine learning to improve suggestions" - Razze - 2018-03-12
(2018-03-12, 00:51)marko.prcac Wrote: HI @Razze !Just to make sure, your aware that this project can't be python only as we lack interfaces into recommendation or well one might say we lack recommendation engine at all.
I have successfully compiled KODI on Ubuntu and I made my hello world add-on today. I am now on my way with producing a DEMO of my idea as an add-on. Is this something that would be interesting for you to see or do you have any other suggestions for further steps?
BR, Marko
So while doing stuff in a kodi addon is nice, it won't help us at all.
We can create models with python as long as we can teach the c++ part to understand those.
Sorry if you know that, but there are so many potential gsoc students, that I'm loosing my overview.
RE: Interested in GSOC project "Using machine learning to improve suggestions" - marko.prcac - 2018-03-22
Hi @Razze !
I understand your point and the task. I have finally found some time to finish my demo: https://github.com/marexv/kodiRecommenderDemo ... I will host it somewhere tomorrow because Heroku wont let me run if off sqllitedb. Demo is intended to be a proof of concept, boards are base only on genre, no tags, no ratings, no ML...
I am running pretty late with my proposal and since you are busy just one question!
Do you want to recommend only movies from users library or from some wider set?
BR! Marko
RE: Interested in GSOC project "Using machine learning to improve suggestions" - docwra - 2018-03-22
If anyone needs help with actual data for this project, I run an open metadata site for music and have many different data points available such as peoples ratings, likes and similar artists, themes, styles and even speed of music.
There is also another nice source called listenbrainz if you like BIG listen data
They have an API that outputs standard JSON and is really easy to use.Lots of cool things could be done with a Standard Kodi Library, history table and some online sources such as these, music AI is an area that is going to be big in the future.
EDIT: For video Trakt.tv is a great resource for TV watch data
RE: Interested in GSOC project "Using machine learning to improve suggestions" - Razze - 2018-03-22
(2018-03-22, 04:52)marko.prcac Wrote: Hi @Razze !I would focus on local as a start. As it won't provide much benefit to recommend stuff, that I can't watch. And we're simply not far enough in our on demand sources in libarary story.
I understand your point and the task. I have finally found some time to finish my demo: https://github.com/marexv/kodiRecommenderDemo ... I will host it somewhere tomorrow because Heroku wont let me run if off sqllitedb. Demo is intended to be a proof of concept, boards are base only on genre, no tags, no ratings, no ML...
I am running pretty late with my proposal and since you are busy just one question!
Do you want to recommend only movies from users library or from some wider set?
BR! Marko
RE: Interested in GSOC project "Using machine learning to improve suggestions" - marko.prcac - 2018-03-23
Local makes sense.
@docwra I will definitely checkout all resources that you shared. Thank you!
Here is hosted version of my demo: http://markoprcac.pythonanywhere.com/ it's something...but its not as good as it can be due to hosting limitations. I had to limit length and number of random walks too much.
BR Marko!
RE: Interested in GSOC project "Using machine learning to improve suggestions" - marko.prcac - 2018-04-15
Hi @Razze!
I've finally had some time to tweak my recommender a bit. Mainly I have implemented a better graph structure so that it now runs a lot faster on hosted site, however it still lacks actors data (next time I'll add those).
My current graph implementation takes up 17MB in JSON file. What is upper limit for KODI plugins size? I'm thinking about writing one regardless of GSOC.
BR, Marko
RE: Interested in GSOC project "Using machine learning to improve suggestions" - Razze - 2018-04-15
I actually don't think we have a limit? At least I can't remember it. Please see the docs in our wiki.