I have recently been tasked with image segmentation of very small granular objects. I asked around, and found that Segmenteverygrain was the model that I needed to use.
I had used a script made to QC the images and come up with labels and masks for each respective image. My question now is, how would I use these masks, labels, and images in order to train a UNet model, which I could then use to better detect images of my choosing?
MODULE HERE: https://github.com/zsylvester/segmenteverygrain