Proposed standardization of the data frame

[iambaim]

I would like to propose new names and types for the annotations in the data frame format:

{
    'ping_time': 'datetime64[ns]',
    'mask_depth_upper': 'float64',
    'mask_depth_lower': 'float64',
    'priority': 'int64',
    'acoustic_category': 'string',
    'proportion': 'float64',
    'object_id': 'string',
    'channel_id': 'string'
}

The above based on the suggestions from @ingeke and myself . What do you think @sindrevatnehol, @albao11 and @ingridut? I plan to propose this to the bigger ICES forum afterward.

Thanks!

[sindrevatnehol]

Hi. I just had a quick look at https://github.com/ices-publications/SONAR-netCDF4/tree/master/docs

the name "ping_time" is consistent with the netcdf time name.

I think this can be posted to the ICES community. There is a posibility to use "backscatter_category" or "Sa_category" instead of "acoustic_category", and also "frequency_id" insted of "channel_id". But such discussion fit better in ICES.

[ingridut]

At NR, we already need to change the annotation format from a dataframe to a matrix in order to the training. The original names and data types do not really matter as long as we can easily identify what each name represent.

I agree with Sindre that "frequency_id" may be a better choice than "channel_id".