Key evaluation updates #396
Conversation
… each error category so new files have been added to the detection folder.
…aluation class uses exactly the same error_scores
Codecov Report
@@ Coverage Diff @@
## master #396 +/- ##
==========================================
+ Coverage 94.6% 94.65% +0.04%
==========================================
Files 93 93
Lines 15078 15197 +119
==========================================
+ Hits 14265 14384 +119
Misses 813 813
Continue to review full report at Codecov.
|
|
Hi @fdlm (or @superbock), Would you mind having a look at this PR ? It should be quick :) It does preserve previous behaviour and adds functionalities that can be useful to other researchers interested in musical key detection evaluation. If there are changes you think are needed I would be happy to discuss. Cheers, Thanks ! |
madmom/evaluation/key.py
Outdated
| 'Relative: {:.3f} Parallel: {:.3f} Other: {:.3f}'.format( | ||
| self.name, self.weighted, self.correct, self.fifth, | ||
| self.relative, self.parallel, self.other)) | ||
| 'Relative: {:.3f} Relative of Fifth: {:.3f} ' |
There was a problem hiding this comment.
We should only report "Relative of Fifth" here if it is actually present in all the evaluation objects.
|
What's the status of this PR? I'm preparing v0.17 and wonder if this is ready to be included. |
|
I guess I have missed the v0.17 release boat, sorry ... it should be done today (Montreal time zone). |
|
No hurry, there's still some more time until the release will be ready. I was just wondering abot the current status. |
|
OK, cool. I'll still try to aim for finishing this today :)
Bertrand Scherrer, Ph.D.
Audio Research Scientist
LANDR Audio Inc.
160 St Viateur East, Suite 809
Montreal, QC, Canada, H2T 1A8
www.landr.com
…On Fri, Oct 25, 2019 at 9:30 AM Sebastian Böck ***@***.***> wrote:
No hurry, there's still some more time until the release will be ready. I
was just wondering abot the current status.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#396?email_source=notifications&email_token=ADPHIF3WLBRYKJIUY3JPUY3QQLYGNA5CNFSM4GHRDVH2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOECILKMI#issuecomment-546354481>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADPHIF562DRCPFOS2IBW2E3QQLYGNANCNFSM4GHRDVHQ>
.
|
…is actually present in evaluation objects
|
@superbock and @fdlm all should be good now. |
|
Ok, v0.17 was delayed a bit ;), but I'm currently working on getting a new version out, so I'd like to check back on this PR and its current status. @bscherrer can you comment on the two comments above? Would be really nice to have this in! |
|
Hi @superbock, I will get this done this week or next. Is that OK timewise ? |
…ding relative of fifth error score
|
You have successfully added a new CodeQL configuration |
separate identification of key error category and score computation: error_scores are now a class variable of the KeyEvaluation class
addition of a new type of error category: relative_of_fifth that plays well with the strict_fifth option
additional documentation for the error_type function
updates to the key evaluation test code and data files to account for the separation of the error category determination and the score to attribute to a given key error
This pull request fixes #395 .