Bug on DATA_CHANNELS Ordering

[NovaBro]

In the documentation for instance segmentation, it states that:
"
In practice, PROBLEM.INSTANCE_SEG.DATA_CHANNELS is a list (or sequence) of these channel codes (e.g. ['F', 'C', 'Db'] or ['F', 'C', 'H', 'V']). The order of the codes defines the order of the output channels of the network.
"
https://biapy.readthedocs.io/en/latest/workflows/instance_segmentation.html

In my YAML configuration, I had set it to :
DATA_CHANNELS: ['F', 'P', 'Dc', 'Dn']

But in the generated labels folder, I get
fisbe/biapy-channel-scale/train/label_F.erosion-0.dilation-0_Dc.type-centroid.norm-True.mask_values-True_Dn.closing_size-3.norm-True.mask_values-True.decline_power-3_P.type-skeleton.dilation-2.erosion-0
Which has channel ordering F, Dc, Dn, P

This caused some confusion, when viewing the labels generated, which I believe is in the ordering F, Dc, Dn, P.
Rather than the order I set in my YAML ['F', 'P', 'Dc', 'Dn'].

[NovaBro]

I tried using cursor to verify this, and said that DATA_CHANNEL_WEIGHTS is not effected by this and is still matched correctly, but I thought I would make this ticket just in case.

[danifranco]

Hi,

Thanks for reporting the issue! You are right, the channels were reordered and there was a bug in the code that now is corrected: 3b09a0b . Pull the changes and try it again please.

Regarding your task, if you provide more details of what you are doing, a few images maybe, I will be more than happy to guide you a bit with the configuration. For example, if you just started using BiaPy I will go with other representations, like F + C , as a starting point and build upon afterwards.