Looking to get some help with annotations.
- Project Type: Gain Detection
- Model Type: Roboflow 3.0 Instance Segmentation (Accurate)
- Checkpoint:COCOs-seg
I have a segmentation model to detect rice and gather the size of each grain using the points I get back in JSON. The classification for size is done on the backend of my project, so it only detects “RICE”.
Now I’m working on discovering if a piece of rice is damaged. Not damaged as in shape but damaged as in burnt / harsh discoloration. I’ll provide an example photo.
My question is what would be the best way to go about this?
Segment the whole grain and label them “RICE-DMG”? That doesn’t seem to be the best option. I could just segment the part of the grain that are damaged but that doesn’t feel right either.
My only other thought is to just detect all grains as “RICE” then feed the individual grains through ChatGPT or equivalent via workflows and ask it if the grains show signs or burning or discoloration?
Any help would be much appreciated!
Sample photo:
