facebookresearch/Mask2Former

Health v2 · maintenance only

Flags: no_releases archived

Adoption not part of the score

3416 stars · 528 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Mask2Former is the official PyTorch implementation of the CVPR 2022 paper 'Masked-attention Mask Transformer for Universal Image Segmentation' from Meta AI Research. It provides a single transformer architecture that handles panoptic, instance, semantic, and video instance segmentation, with pretrained models for major datasets like COCO, ADE20K, Cityscapes, and Mapillary Vistas.

Use cases

• perform panoptic segmentation on images
• run instance segmentation with a transformer model
• semantic segmentation on Cityscapes or ADE20K
• video instance segmentation
• fine-tune a state-of-the-art segmentation model on my own dataset
• download pretrained segmentation checkpoints
• compare segmentation baselines across datasets

When to choose

• you need one architecture covering panoptic, instance, and semantic segmentation
• you want strong pretrained baselines on COCO, ADE20K, Cityscapes, or Mapillary Vistas
• you are doing research on segmentation transformers and want a reproducible reference implementation
• you need video instance segmentation support

When to avoid

• you need a lightweight model for real-time or edge deployment
• you want a maintained production library with frequent updates - the repo is in maintenance mode
• you need segmentation for domains without supported pretrained weights and lack training resources
• you prefer a simple API over research-style codebases built on Detectron2

Facets

library · maturity maintenance

computer-vision image-processing machine-learning deep-learning computer-vision image-processing deep-learning machine-learning python cross-platform image-segmentation panoptic-segmentation instance-segmentation semantic-segmentation transformer cvpr-2022 research-code pytorch video-instance-segmentation model-zoo linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebookresearch/Mask2Former")