Component Labeling for a food detection model

Describe your question/issue here! (delete this when you post)

I am currently trying to train a yolo26s model for plated food detection in a restaurant setting. I built classes for the menu item name and annotations around the whole dish with successful map scores above 90%. I am having trouble with components on the plate, is there an easier way to help the model detect components on the plate and train it to decide if those components are present or not. I see the polygon option, how would that work, do I create sub-classes and what should I name them to not confuse the model with the main menu item classes.

Hi @Mike_Spain ,

Could you share some results? And have you tried using some VLMs as well? Likely those could work out of the box, or with minimal fine-tuning. Could you share your workspace and let us know if we have perm to look into it? We can also hop on a quick call, I’d be interested in hearing more about the application - let me know at erik@roboflow.com .

Thanks, Erik

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.