arXiv 2112.01527
Masked-attention Mask Transformer for Universal Image Segmentation
By Bowen Cheng, Ishan Misra, et al.
Published 2021-12-02
Discussion
Read the public discussion and references gathered around this paper.
Image segmentation is about grouping pixels with different semantics, e.g., category or instance membership, where each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing specialized architectures for each task. We present Masked-attention Mask Transformer (Mask2Former), a new architecture capable of addressing any image segmentation task (panoptic…