datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
smoking_classificationmai-smoke
MAI real-photo versus generated-image atlas v2 pilot
A paired pilot in which pinned public-domain real photographs are deterministically captioned and rendered by every generator family.
This repository is produced by the MAI provenance-first dataset pipeline.
Normalized images and their typed metadata are embedded in Parquet; byte-identical
originals, provenance receipts, checksums, generator settings, and
semantic-group relationships are retained for audit.
Dataset-specific… See the full description on the dataset page: https://huggingface.co/datasets/kennethnym/mai-smoke.