Unable to perform intersection segmentation(time&frequency) on audio files

[duanhuiran]

It appears that the MeGraS system currently does not support intersection segmentation for audio files, where segmentation is applied both in the time and frequency domains simultaneously (e.g., segmenting a portion of background music from whole audio).

Reproduction steps:

1. Upload an audio file.
2. Attempt an intersection segmentation, for example: http://localhost:8080/[audio_object_id]/segment/time/300-5000/frequency/1000-4000

Result: 404 Not Found (no error output from the backend).

Could you confirm if this feature has not yet been implemented in MeGraS? If it hasn’t been implemented, I may need to reconsider the segmentation of background music.

[duanhuiran]

@lucaro

[lucaro]

Your syntax is wrong. Try http://localhost:8080/[audio_object_id]/segment/time/300-5000/segment/frequency/1000-4000

[duanhuiran]

emm... I don't think so.

Please see syntax in thesis:

And the result of http://localhost:8080/[audio_object_id]/segment/time/300-5000/segment/frequency/1000-4000:

[lucaro]

The thesis is not the source of truth; the implementation is. The fact that you get a segmentation error rather than a 404 also tells you that this is the way megras expects this to be specified. That you get an 'invalid segmentation' error is a problem, though. I'll have a look why that might be

[lucaro]

There was a segmentation definition encoding problem that generated invalid URLs in the process. I fixed that in https://github.com/lucaro/MeGraS/commit/0bd62c1f4848ca23d1164d35916eaaaddf8645f5. After that fix, localhost:8080/[...objectId...]/segment/time/300-5000/segment/frequency/1000-4000 works for me.