SAM 3 & Open-Vocabulary Segmentation

Give a model a text prompt and an image and get masks for every matching object. SAM 3 made that a single forward pass.