---
license:
- other
- cc-by-4.0
- cc-by-2.0
- cc-by-nc-2.0
- cc-by-nc-sa-2.0
configs:
- config_name: lvis
  data_files:
  - split: train
    path: lvis/train-*
  - split: validation
    path: lvis/validation-*
- config_name: panoptic
  data_files:
  - split: train
    path: panoptic/train-*
  - split: validation
    path: panoptic/validation-*
source_datasets:
- winvoker/lvis
- srishti-kaushik/COCO-2017
---

# coco-regions

LVIS v1 (1,203 classes) and COCO Panoptic (133), sharing one COCO image cache.

Image licenses remain per-image; NoDerivatives images are excluded from
rendered mirrors. Only COCO train2017 images can enter training; evaluation
uses the intersection of native annotation validation and COCO val2017.
This deliberately excludes conflicting LVIS/COCO image partitions, rather
than silently relabeling them. No unlabeled test images are included.

Original data: [winvoker/lvis](https://huggingface.co/datasets/winvoker/lvis), [srishti-kaushik/COCO-2017](https://huggingface.co/datasets/srishti-kaushik/COCO-2017). Repackaged as parquet for [tasksource](https://github.com/sileod/tasksource) by [scripts/repackage_dataset/regions.py](https://github.com/sileod/tasksource/blob/main/scripts/repackage_dataset/regions.py#L165).

[Conversion provenance](provenance.json): source revisions, hashes, configuration and exclusions.

## Release scope

Bounded samples of the eligible native annotations; full classification ontologies are retained.

- lvis: train: 99,959, validation: 100.
- panoptic: train: 100,000, validation: 100.

See [provenance.json](provenance.json) for source pins, selection seeds, rendering configuration and exclusions.