Training Segmentation Model with Custom Labels in MonaILabel

I am looking to train a segmentation model using MonaILabel. However, instead of using the pre-existing labels such as (‘kidney_left’, ‘kidney_right’, etc.), I want to use custom labels (‘label1’, ‘label2’, ‘label3’).

My dataset already has segmentation maps defined.(had 3 labels) Could you guide me on how I can train the provided segmentation model in MonaILabel with these new labels?

If there are any related documents or resources that could help me understand this process better, please do share them.

There are several good training videos of @diazandr3s available.
Here is one at the timeframe where it is shown how to set your own labels.
As MONAILabel is always expanding and prone to change, you might need to re-ask.

Assuming you are planning to use, you need to change these lines to match your labels. However, I would advise against changing them to label_1, label_2, but instead use proper names of what they are…

1 Like