Segment Anything: A Promptable Foundation Model for Image Masks
Kirillov et al. built a ViT image encoder, a prompt encoder, and a mask decoder trained on a billion masks. Why promptable segmentation is a product interface, not just a COCO score.
3 min read