Fine-tuning SAM 2 to Output Segmentation Masks with Class Names

I am working on a project that involves using SAM 2 for image segmentation, and I would like to explore the possibility of fine-tuning the model to output segmentation masks along with their corresponding class names. The goal is to not only obtain the segmentation masks but also associate each segmented object with its respective class name.

Start by reviewing the repository's inference code and example notebooks to understand how SAM 2 currently produces segmentation masks. Determine what fine-tuning workflow is needed and how class names would be represented alongside each mask; done means a documented approach or working example that produces both outputs.

Written by the indexing model from the issue text.