Clarification for the use of the automatic horizontal noise band removed

[SnowySnowySnowy]

I saw that in the latest DeepSqueak release there's automatic removal of horizontal noise bands. I have some audiofiles that have this (as shown below). My question is how I get this to work? What network should be used? When I just tried standard call detection it is still including the noise band as calls.

[DrCoffey]

hey @SnowySnowySnowy . The horizontal noise removal is baked into the create_training_images and the new netwoks (RatYoloR2, etc). It won't remove the lines from the actual audio. The code only removes consistent horizontal lines, so i don't think it would work on that noise. It is always better to find and remove the source of the noise than fix post hoc.

Actual code is:

% Calculate Horizontal Noise
rowFifth = prctile(p,5,2); % Lowest Fifth Percentile of Eack Row
highLows = prctile(rowFifth,90); % Top 10% of lowest 5th Percentiles
sDev = std(p,0,2); % Standard Deviation of Rows
lowDevs = prctile(sDev,10); % Lowest 10% of Standard Deviations
idx = rowFifth>highLows & sDev<lowDevs; % Rows with very high fifth percentile brightness and low standard deviation - AKA Solid Lines...

% Remove Horizontal Noise from lines, add back the fifth percentile of the
% entire spectrogram
p(idx,:) = p(idx,:) - prctile(p(idx,:),5,2) + repmat(prctile(p,5,"all"),sum(idx),1);