Ml5-library: Where to put entire datasets
@cvalenzuela pointed out that data currently lives in /docs/assets/, though it looks like /data mostly holds trained output or something and /img looks mostly like singular images. How does everyone feel about adding /datasets under /assets to hold entire datasets?
handav
All 12 comments
Since we are moving to a new github handle, maybe having a repo just for datasets can be useful
cvalenzuela
on 5 Mar 2018
I'm game for a datasets repo! However, I wonder if having them in the "examples" repo would make sense -- do we imagine having datasets that don't pair with an example?
shiffman
on 6 Mar 2018
Maybe a repo with Training and Datasets would make more sense? moving the current /training folder to a separate repo makes sense to me
cvalenzuela
on 6 Mar 2018
Also if you put datasets in a repo they could be included as a dependency in non-ML5 projects.
dariusk
on 6 Mar 2018
Sold! Should we call it:
- ml5-data
- ml5-training-data
- ml5-datasets
- ml5-data-and-training
??
shiffman
on 6 Mar 2018
+1 for ml5-data-and-training
cvalenzuela
on 6 Mar 2018
I like that too! Or ml5-training-data, though ml5-data-and-training is a
little bit clearer.
On Tue, Mar 6, 2018 at 5:14 PM, Cristóbal Valenzuela <
[email protected]> wrote:
+1 for ml5-data-and-training
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/ml5js/ml5-library/issues/85#issuecomment-370948168,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABUjXMpnS1fyTNvuBs9Q0sXoKvcg2iyVks5tbwpBgaJpZM4Sdyw5
.
handav
on 6 Mar 2018
The repo is live! (@handav you should have admin privileges but feel free to add datasets via pull requests so we get notified. For simple and quick changes like README updates, etc. I think writing directly to master is a-ok! And happy to sit down and go through this with you of course.)
shiffman
on 7 Mar 2018
(Will leave this issue open until we move some of the data itself.)
shiffman
on 7 Mar 2018
All data has been moved!
cvalenzuela
on 13 Mar 2018
I still have to put my dataset up though! I just have to weed out the ones that people tagged as being flawed, and then I'll get it up. Aiming to do that and the music LSTM by Thursday.
handav
on 13 Mar 2018
cool! we can follow this on #56 too.
cvalenzuela
on 13 Mar 2018
Related issues
lastnod
·
4Comments
K1ngjulien
·
6Comments
shiffman
·
5Comments
yining1023
·
9Comments
oveddan
·
9Comments
Most helpful comment
Maybe a repo with Training and Datasets would make more sense? moving the current
/trainingfolder to a separate repo makes sense to me