Universal Image Segmentation via Mask2Former and OneFormer
Original title: Universal Image Segmentation with Mask2Former and OneFormer
35AI Score
Mask2Former and OneFormer demonstrate a unified architecture for semantic, instance, and panoptic segmentation. The research simplifies the segmentation pipeline via masked attention, achieving universality across tasks.
Related Entities
Full Report
Universal Image Segmentation with Mask2Former and OneFormer