SwatPhonLab / UltraTrace

A Free/Open-Source tool for manual annotation of Ultrasound Tongue Imaging data.
GNU General Public License v3.0
10 stars 5 forks source link

a few issues when dealing with ULT data #128

Open jonorthwash opened 4 years ago

jonorthwash commented 4 years ago

Apparent regressions:

Other stuff:

These can be made separate issues if desired.

jonorthwash commented 4 years ago
jonorthwash commented 4 years ago
jonorthwash commented 4 years ago
  • [x] Figure out what's up with the extra set of traces called "null" being added.

This apparently was left over from the alignment issue. Deleting metadata.json and running again seems to've fixed it.

jonorthwash commented 4 years ago
mr-martian commented 4 years ago
  • [ ] Are frames being overwritten?
INFO:   Line 21 of SPLINES.dat imported as File010 frame 99
INFO:   Line 22 of SPLINES.dat imported as File010 frame 100
INFO:   Line 23 of SPLINES.dat imported as File010 frame 99
INFO:   Line 24 of SPLINES.dat imported as File010 frame 100
INFO:   Line 25 of SPLINES.dat imported as File010 frame 99
INFO:   Line 26 of SPLINES.dat imported as File010 frame 100

Yes, because that SPLINES.dat has 3 consecutive pairs of lines with the same date. They have different annotations though, perhaps it should try to load the textgrid and check whether the label matches?

  • [ ] Why are some lines unable to be imported?
WARN:   Unable to import line 71 of SPLINES.dat
WARN:   Unable to import line 72 of SPLINES.dat
WARN:   Unable to import line 73 of SPLINES.dat

Because the File030.ult's mime type comes up as application/x-dosexec, so it's ignored and then the importer won't be able to compute frame times so it skips File030. Fixed in 7c6fec5.

jonorthwash commented 4 years ago

Yes, because that SPLINES.dat has 3 consecutive pairs of lines with the same date.

That in and of itself isn't a problem. I guess the issue here is the second field, "time of sample in recording". These lines, besides their annotation label, appear to be legit repeats. I wonder why the AAA export did that....

They have different annotations though, perhaps it should try to load the textgrid and check whether the label matches?

Yeah, it would be good to try to match the label to whatever's in the TextGrid at that time index. Normally throw a warning for non-matches, and when there are conflicts, go with either the correct one or the last(?), or both if there are multiple repeats (like their are here).

As long as we load the right data for this file, it's not very important.