About Semantic Segmentation

[ArmandAlbert]

Hello, I want to ask how to achieve the semantic segmentation effect described in the article images, such as identifying a person, cow, car?

[lkeab]

Hi thanks for the question. For example, on the COCO dataset, we use a SOTA detector FocalNet-DINO trained on the COCO dataset as our box prompt generator to produce mask. The box prompt already contains the category information.

[alexwong-16]

I wonder whether SAM-HQ still has the ability to perform zero-shot labeling by using detector for semantic segmentation.

[Ghy1209]

Hi thanks for the question. For example, on the COCO dataset, we use a SOTA detector FocalNet-DINO trained on the COCO dataset as our box prompt generator to produce mask. The box prompt already contains the category information.

Hi . I want to know when you use the box prompt, if only boxes greater than the threshold are used as input？