#metabrainz

Shocker. Vueling flight home is... Delayed.

Hi there.. I'm interested in "ListenBrainz: Statistics" project . Please help me getting started or any warm-up tasks?

At least I got reminded of https://www.youtube.com/watch?v=jVFZLXtq0nE :p

"Caught 'em up in the web of the world and it's wide / The YouTubes has more science than Bill Nye the Science Guy" :D

armalcolite: Can you tell me your name? :p

Freso: we should probably look at the old forum stickies and see if some should be imported

Sure.

reosarevok: I should be able to do the renames later today/tomorrow.

:)

But I have learned not to give ETAs if they're not strictly needed :p

But first, back into sun and more sanity than this beaurocratic crap city. :-)

Man. I can't wait for June. Borrowed a replacement laptop, but it's so slow. ;_; I've put in 4 GB of RAM, but it seems like it only recognises 2. And the up key doesn't work.

And the mousepad moves the mouse all over the place.

Except now it does. Wtf.

I added some "Getting started" subsections to the ideas page, after discussion with Gentlecat last night

suggesting what people can do with the AB and LB servers to get started - I was getting the feeling that our blanket "set it up and play with it" was a little too vague

"set it up" isn't, "play with it" is. :)

but I even think "set up LB" is kind of vague without saying why

yes, I know "implement LB-1" or similar is a good reason, but I think explicit is better than implicit

anyway, I agree. if we can get people coming in saying "hey, I set this up and tried some things, what should I do next" I think that's a lot better than "hey, I want to work on x, what should I do?"

Agreed.

so, let's see if these comments change the tone of questions around MB and LB

s/MB/AB/

*B :)

[AL]B

(A|L)B

well. that's the same as [AL] :)

I guess it depends what grammar you use

(A|L)+B

I'm trying to clean up GettingStarted to include some less-MBS specific stuff

Freso: is the submission template on the wiki?

GSoC application? Yes.

It'd be nice to expose students to it before they apply

where?

Summer of Code/Application Template or something like that. It's linked from the blog post at least.

OK, I'll add it to some other wiki pages too

:)

Thank you alastairp.

you're welcome, head of community Freso

;)

:p

Hi guys, I'm interested in contributing to me MetaBrainz via gsoc this year

So I came to say 'Hi'

:)

hello

hi

Concretely, I'd like to work on AcousticBrainz, because it includes (as I understood it) lightweight machine learning experience

"Lightweight"...

:)

But hi :)

we currently use SVM on AcousticBrainz

it's described here: http://essentia.upf.edu/documentation/streaming...

alastairp: thanks!

Freso: "lightweight" as in "It would suit someone with experience or an interest in machine learning algorithms, though the majority of the project will probably involve creating infrastructure around our existing algorithms." :)

we don't have any concrete proposals specifically around the machine learning algorithms

right, as you've just seen

however if you wanted to propose something we could probably make it fit

new algorithms to see how they compare to SMV? Random Forest? Deep learning?

I'll check the link above and AcousticBrainz docs and see if anything comes to mind

Freso: wouldn't it discourage people from discussing a project if there is no category for it? doesn't seem like the right approach to me

alastairp: how do Python and Flask come into play with AcousticBrainz since they're listed on the ideas page?

the server is written in flask

there's a audio feature extractor written in C++ which performs the signal processing and outputs json

then the webserver does stuff with it

"stuff"

discouse is pretty buggy

how do I turn of smilies in my post on discourse?

Hi ...I am Sandeep bharti currently pursuing my under graduate program in computer science and I am interested in the Listenbrainz:Statistics project as listed in the project ideas for GSoC 2016. Can anyone help me in how to get started ? Thanks in advance :)

hey. one way to start is to try and install ListenBrainz to get familiar with some things

alastairp: is there a git repo for the webserver code?

nevermind, found it!

sandy_: there is a "Getting started" section on the Ideas page. You should see if you can follow that at first

cetko: cool

okay it looks cool. I will follow that for now, thanks :)

kind of weird that h3 and h4 look exactly the same in the wiki

Gentlecat: yeah

I see an issue notice about the statistics chart on the accousticbrainz webpage, what kind of issue?

http://acousticbrainz.org/statistics-graph see all the wobbles?

http://beta.acousticbrainz.org/statistics-graph no wobbles - issue fixed!

oh, I thought that's supposed to be there :)

finally caught up \o/

a side note, I've recently dabbled with real-time databases (InfluxDB) for pageview analytics, would that be of any interest for the statistics part of acousticbrainz?

AB stats aren't really that critical

and their scale is quite small. we have a few rows per hour storing counts, and that's it

yes, but as I understand it, the c++ submission code outputs json, are those raw time-series data?

or aggregated?

ah, no. statistics is just a time series of the number of data submissions we have

the output of the feature extractor is not time series

alastairp: Replied to your comments on my ideas. Please have a look whenever possible and we can discuss it further.

we take a mean (and other statistics) over the entire contents of the audio file

however, we are looking at storing complete time-series data in the future as well (we call it frame-level data)

I wonder if a real-time db would be the kind of thing that we can use to store this frame level data

alastairp: how far in the future? this summer? :)

perhaps

the extractor generates the data with a configuration option switched on

we currently store json in postgres, in a jsonb field

however once we generate the extended data it's too large to store in postgres (10MB/minute)

we don't actually know how to store it yet. I had a look into protocol buffers, but I'm also interested in looking at other options

what's the extended data?

one sec, let me see if I can find an example

Hi again... sorry I lost my connection then... I'm interested in "ListenBrainz: Statistics" ... So how should I proceed?

in fact, you could probably generate it yourself

have you downloaded streaming_extractor_music ?

himanshu25: take a look at the "Getting started" section for listenbrainz on the wiki

cetko: http://www.dtic.upf.edu/~aporter/ab/roxy.json and http://www.dtic.upf.edu/~aporter/ab/roxy.json_f...

see in the json we have summary information - mean, median, variance, first derivative for some values

whereas in the frames data, it has 1 value for every 1024 frames of audio (1 second = 44100 frames)

so, e.g. the barkbands_crest->mean value is the mean of all of these frames

kartikgupta0909: just pinged you. :p

alastairp: so streaming_extractor_music generates both summary and frames data?

cetko: right

if you look at the usage (run with no arguments) you see that there is an argument for a profile file

put this in a file and use it as the profile: https://gist.github.com/alastair/df17f5f641c760...