I have a new task I am currently working on involving with some description below:

The forensic laboratory uses a 3D microscope to capture images of fired ammunition components, specifically cartridge cases. These acquired images are then sorted by software for comparison purposes to learn if the cartridges in question came from the same or different firearms, which is information of great value to police investigations.

Masking/coloring is a preparatory function, currently performed manually, which assists the software to identify on the image-specific features of interest on the cartridge case. This manual process is tedious and time-consuming, and automating it would be a significant improvement.

The objective is to create an algorithm to automatically mask cartridge case images.

I have never done a project like this before but I have used CNNs to classify types of photos before but this is a very different task with regards to forensics.

Any approaches or ideas on where I should start and the type of algorithm needed?