Open sindrevatnehol opened 3 years ago
Including the users @ingeke , @ingridut, @albao11 in the discussion.
We already alter the annotation labels in our current code for training purposes, and we don't require the original labels to be of any specific format or type. Our main concern is that we need to be certain which species each original label belong to, and the categories from the ICES vocabulary seem quite intuitive to me.
Data from IMR should be ok as we use the same reference per species. But including data from other institutes/nations, the reference may be different. Switching, and only allowing ICES vocabulary, would serve as a common reference, avoiding any confusion of the species.
We need to copy the reference table to git.
LSSS labels different acoustic categories using integer, where a reference has been developed at IMR. I.e. for herring the acoustic category is labeled as '12', but the user may define a different reference.
To make annotation data more readable, we propose to convert LSSS labels to ICES vocabulary.
The different acoustic categories is available here http://vocab.ices.dk/?theme=4 via AC_SaCategory.
This conversion will cause the acoustic_category of i.e. herring to become 'HER' instead of 12.
Another option is to expand the output table with another column, where one indicate LSSS/software notation and the other the ICES vocabulary.
Does anyone have any thoughts regarding this?
Does anyone have any thoughts regarding this?