You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Feature Extractor audio clip length and noise removal clip start and length are all now in milliseconds instead of samples. Keep note of this if you have configs with an instance of the Feature Extractor saved into them!
Fixed major bug in Feature Extractor where the end sample was input into the sample retrieval function instead of the clip duration when using a fixed clip length.
Minor work-around in the Feature Extractor for a bug that causes some call clusters to be null.
Fixed oversight in Training Set Builder where Feature Extractor settings were not read back in.
Exporting full classifier results now takes significantly less time.
Default file names for classifier output slightly changed.
Fixed some bugs involving the Feature Extractor's success/failure/ignore/pending counters.
Fixed oversight when exporting full classifier results where feature names were not listed, and also added plugin version, added hydrophone location, and fixed time zone miscalculation.
Fixed oversight in WMNT where changes weren't logged and undo button didn't work when double-click editing an individual cell.
Added loading bar in WMNT for when database entries are being loaded in, as premature editing was causing some exceptions.