# Unified Multi-task Architecture

Mask2Former is a general-purpose image segmentation model based on Transformer architecture, capable of handling instance segmentation, semantic segmentation, and panoptic segmentation tasks.

Mask2Former model trained on the ADE20k panoptic segmentation dataset using a Swin large backbone network, employing a unified paradigm to handle instance segmentation, semantic segmentation, and panoptic segmentation tasks.

Mask2Former model based on Swin backbone network, specifically optimized and trained for panoptic segmentation tasks on the Cityscapes dataset

Mask2Former model based on Swin-Tiny backbone, optimized for Cityscapes panoptic segmentation tasks

Mask2Former is a Transformer-based unified image segmentation model supporting instance segmentation, semantic segmentation, and panoptic segmentation tasks, utilizing masked attention mechanism to enhance performance