Background Regions overlap with other regions

[Joetown94]

Hi,
I am looking to complete full image segmentation, our workflow right now is to annotate a series of specific regions within a 10m x 10m image (corals, sponges, etc.), but then annotate the remaining parts of the image as "bare substrate" once all other pieces are labeled. Is there a good workflow for this?

When using the "Create Background Region using WA" tool, the region created overlapped with the other annotated regions. The other tools to separate them require pairwise selection and removal, which would not be feasible (>500 other regions).

What would be the best way to make a region of the background that excludes existing regions?

[Jordan-Pierce]

Is "bare substrate" a class or are you saying that everything not labeled is bare substrate, and is considered the background? I.e., do you want the model to actually learn and predict bare substrate? If you want the model to learn / predict bare substrate, then definitely assign it a class label, otherwise keep it as the background class (either 0 or 255, I forget).

If everything is already labeled and the only thing remaining is bare substrate, and you need those pixels labeled as "bare substrate" you can just do that in python (or maybe even paint). I'm imagining that everything is labeled except bare substrate: pick out that pixel in the image, find the corresponding pixel in the segmentation mask, identify the current value in the mask (probably 0 or 255), and reassign all values that are 0 / 255 to the bare substrate class value (not 0 / 255)