LibreEoMT third-party notice
============================

LibreEoMT integrates EoMT (Encoder-only Mask Transformer) for semantic,
instance, and panoptic segmentation.

Upstream:
  Source: https://github.com/tue-mps/eomt
  License: MIT
  Copyright: Mobile Perception Systems Lab at TU/e and contributors

Runtime reference:
  Source: https://github.com/huggingface/transformers
  License: Apache-2.0
  Copyright: The HuggingFace Team and Mobile Perception Systems Lab at TU/e

Backbone:
  DINOv2 is Apache-2.0. LibreEoMT ships only DINOv2-based EoMT checkpoints,
  sizes s/b/l. DINOv3 EoMT variants are excluded because they depend on
  gated non-commercial DINOv3 weights.

Weights (all MIT licensed by upstream):
  Semantic (ADE20K 150-class, 512px, size l):
    tue-mps/ade20k_semantic_eomt_large_512
  Instance segmentation (COCO 80-class, size l at 640px and 1280px; upstream
  ships no DINOv2 instance checkpoints at s/b):
    tue-mps/coco_instance_eomt_large_640
    tue-mps/coco_instance_eomt_large_1280
  Panoptic (COCO 133-class, 80 things + 53 stuff, 640px, sizes s/b/l),
  first-class task="panoptic" checkpoints carrying thing_class_ids metadata:
    tue-mps/coco_panoptic_eomt_{small,base}_640_2x
    tue-mps/coco_panoptic_eomt_large_640

  ADE20K and COCO are research datasets; users remain responsible for
  dataset-license compliance when validating or fine-tuning.
