Heya lucifer and monkey - have had this question on Mastodon and the Forums now, so I’m going to bug you about it: When we re-run YIM 2024 will/can we also rerun previous years YIM for users?
I’ll also put up some socials now to entice people to signup/import last.fm profiles (saying that we will rerun YIM next week)
<aerozol[m]> "Heya lucifer and monkey - have..." <- it should be possible for 2023 but not for years before that because the code and design has changed a lot.
<OmSantoshSuneri[> "Do we have GSOC projects for 202..." <- GSoC for 2025 and projects have not been annouced yet.
OmSantoshSuneri[
lucifer[m]: Is it in doubt that GSOC 2025 will be held or not like
lucifer[m]
Om Santosh Suneri: no idea, its totally upto gsoc/google admins. but the timeline will definitely be delayed if/when its announced.
d4rk-ph0enix has quit
d4rk-ph0enix joined the channel
OmSantoshSuneri[
<lucifer[m]> "Om Santosh Suneri: no idea..." <- ohh fine i was just wondering that are they organising gsoc this year or not
Thanks bitmap, but it didn't really change anything
d4rk-ph0enix has quit
d4rk-ph0enix joined the channel
BrainzGit
[listenbrainz-android] 14hemang-mishra opened pull request #522 (03main…Fix-MoveLogicToVMInArtistScreen-hemang_mishra): Moved card Link logic to viewModel https://github.com/metabrainz/listenbrainz-andr...
d4rk-ph0enix has quit
d4rk-ph0enix joined the channel
Shreyas[m] joined the channel
Shreyas[m]
Hi jasje,
I was looking at the ArtistsOverviewScreen and considering adding a dropdown feature where users can view the songs of an artist using ListenCardSmall. What are your thoughts on this? If it sounds good, I'll go ahead and open a PR for it.
Shreyas[m] uploaded an image: (64KiB) < https://matrix.chatbrainz.org/_matrix/media/v3/download/matrix.org/akTcefbXVvaWbbeEpnpMKnbu/photo_2025-01-09_16-07-23.jpg >
jasje[m]
Shreyas[m]: Sure, go ahead. I'll review all the PRs on coming weekends :)
Shreyas[m]
jasje[m]: Thanks! I'll get started and submit the PR soon :)
petitminion joined the channel
petitminion has quit
d4rk-ph0enix has quit
d4rk-ph0enix joined the channel
monkey[m]
reosarevok: Hello! I'm looking into that support email you forwarded, some code in LB is breaking because of a date formatted like so: "1993-??-01".
Are partial dates with year+day but no month allowed? If so why?
I looked into the original entry and source for https://musicbrainz.org/release/774d0f07-a554-4... , and the Discogs date in "Jan 1993" so it looks like an error , but I'm surprised that missing-month-format would be valid.
In any case I'll fixed the code in LB. This question is more for my own education
d4rk-ph0enix has quit
d4rk-ph0enix joined the channel
reosarevok[m]
Why not? If we know the year and day but not month that seems like valid data 🤷♂️
petitminion joined the channel
monkey[m]
Not to me :p How do you know the day of the month without knowing the month?
But I guess I am currently looking at it from a date parsing and formatting point of view, so limited.
The browser sure doesn't want to format a date as "The first of who knows which month 1993"
Hi jasje , Hemang here. I have done a few PRs in the past month in the listenbrainz repo. It would be great if you let me know any specific feature or improvement in the app, you would like me to work on.
* the listenbrainz android repo. It
jasje[m]
HemangMishra[m]: Hey Hemang Mishra: It's great to see the contributions! We have two pages missing on profile tab, Playlists and Created for you. You can pick these pages up and implement them as are on the website.
HemangMishra[m]
jasje[m]: Okk thanks a lot. I would love to work on them.
mayhem[m]
jasje: ^^
jasje[m]
mayhem: already handled 🏃♂️
<HemangMishra[m]> "Okk thanks a lot. I would love..." <- Hemang Mishra: Pull in aerozol @aerozol:matrix.org: If you might have doubts regarding designs, but overall, i think the website mobile UI should be a good starting point.
jasje[m]: Hemang Mishra: also, you can go a step ahead and create a figma file which you can share with us and plan your steps
HemangMishra[m]
jasje[m]: Okay, I will do that. Thanks for the guidance.
mayhem: outsidecontext zas was chatting with mayhem about the work I’m doing with essentia to get reliable automated tempo estimation. I’ve been messing around with this for a couple weeks now and I have a pretty good feel for it. Bottom line is that I’m kind of shocked at how reliable it is, not just at getting the pulse train for a track, but the actual correct BPM which also has the time signature baked into (a BPM can
commonly be off by a factor of 2 if you don’t know what note gets the beat). It falls short in 3 main areas, 2 of which I think we shouldn’t vend tempo data for, and a 3rd may be a lack of training on that genre
* that genre 🧵
igold94[m]: It struggles with classical/baroque music which tends to not have a consistent BPM in which case we shouldn’t vend this data, and it also naturally struggles in spoken word tracks like comedy or books being read (also should not be vended). The model uses a soft max on the output so the probability output is quite reliable in identifying that something is likely not a measurable tempo
igold94[m]: Finally it struggles with its confidence primarily in metal genres because they often contain multiple time signature changes, as well as multiple tempos. Despite the fact that it’s not confident in its answer, it’s quite often correct and I haven’t analyzed its multi-tempo outputs. I can play around with that to see if I can get a more logical output for those genres than a single global BPM
igold94[m]: Oh it also struggles a bit with tracks that are in half time but this could also be a lack of training (something we can address)
igold94[m]: Bottom line is that this data is FAR more reliable than Spotify in my experience so I think we should provide tempocnn as a capability in Picard to start collecting this
igold94[m]: I’m going to continue this work and find a reliable means of gathering time signature and move on to some other key data points piecemeal versus something more ambitious like acousticbrainz
igold94[m]: Oh also worth noting that I tested this against 1 track of all of the 6k+ genres identified by Spotify. It performed extremely well across the board with the above caveats
outsidecontext[m
igold94[m]: igold94: This sounds very promising. And I also think a project more limited in scope to gather tempo data might have more success. We had essentia tools bundled with Picard at some point. We could re-add this more or less easily I think. More work would be to put up the submission, but if we have somewhere to submit to that's not too much work either I think.
igold94[m]
outsidecontext[m: Ya the only think to play with is the fact that this is essentia-tensorflow as it’s ML based. I know there can be some platform-specific issues with tensorflow so just something to look out for
igold94[m]: We may wanna discuss schema to future proof it if we do want to allow for multi-tempo identification, as well as messaging to the user in cases where we don’t have confidence in the tempo estimation. Ideally users would verify the tempo before submitting but I’m new hear don’t know if that’s feasible or not
igold94[m]: Could use a voting system if multiple submissions are put in for the same track (eg: mode, avg, etc.)
monkey[m]
igold94[m]: It would be wise to assume a lot of users won't know how to verify or calculate a tempo.
igold94[m]
monkey[m]: Fair. Could include a metronome with Picard 😅. That’s what I was going
igold94[m]: Otherwise we can just capture the confidence as well and let consumers decide if they trust it or not
outsidecontext[m
igold94[m]: I think a voting where all submissions with bpm + confidence are stored, and the service returns the median bpm or such, would work well
vardhan__ has quit
MonkeyPython has quit
MonkeyPython joined the channel
BrainzGit
[listenbrainz-android] 14Shreyassp002 opened pull request #523 (03main…Feat-Dropdown-Songs): Feat: Dropdown songs card for AlbumsOverViewScreen & ArtistsOverviewScreen https://github.com/metabrainz/listenbrainz-andr...
