Datasetpublic

bengaliAI/badlad-results

BaDLAD — public results Score rows for the maintainer-run BaDLAD leaderboard. Field Meaning model_id Hub id or slug model_url Link to weights mask_map COCO mask AP@[.5:.95] on private paper hidden test (higher better) mask_map_by_domain JSON map domain → mask_map bbox_map COCO bbox AP@[.5:.95] (optional) backend Decode stack used by maintainers scorer_commit / decode_commit Git SHAs in BengaliAI/BADLAD evaluated_at ISO date requested_by Who asked, or… See the full description on the dataset page: https://huggingface.co/datasets/bengaliAI/badlad-results.

sourceHugging Facecc-by-4.0updated 2mo agoView on Hugging Face

0likes33downloads

Dataset Card

BaDLAD — public results

Score rows for the maintainer-run BaDLAD leaderboard.

Field	Meaning
`model_id`	Hub id or slug
`model_url`	Link to weights
`mask_map`	COCO mask AP@[.5:.95] on private paper hidden test (higher better)
`mask_map_by_domain`	JSON map domain → mask_map
`bbox_map`	COCO bbox AP@[.5:.95] (optional)
`backend`	Decode stack used by maintainers
`scorer_commit` / `decode_commit`	Git SHAs in `BengaliAI/BADLAD`
`evaluated_at`	ISO date
`requested_by`	Who asked, or `maintainer`
`notes`	Caveats

How eval works

—Hidden test (badlad-test, paper COCO) stays private (no public images / gold / hypotheses).
—Request eval: BengaliAI/BADLAD issues.
—Maintainers decode on closed test and append a row here.
—Display: `bengaliAI/badlad-leaderboard`.

Do not upload prediction files or test labels to this dataset.

Reproducing scores

Maintainer tooling: `BengaliAI/BADLAD/leaderboard` (uv sync, Detectron2 installed separately). Baseline: leaderboard/baselines/2026-07-25-mrcnn-scratch.metrics.json.