bluzen/bluzen-tagger-v1
Image Classification
Multi-label illustration tagger over a 12k-tag vocabulary, trained on the Aurora corpus with a ViT-L backbone.
8 published entries across models, datasets and demos.
Image Classification
Multi-label illustration tagger over a 12k-tag vocabulary, trained on the Aurora corpus with a ViT-L backbone.
Image Classification
Preference-trained aesthetic scorer calibrated on paired human judgements rather than scraped star ratings.
Image Classification
Content gate with published per-class error rates, meant for dataset filtering rather than moderation decisions.
Image Classification
The full Aurora illustration corpus: deduplicated images with tag, caption and quality annotations.
Image Classification
A 1.2M-row quality-filtered slice of Aurora. The subset most of our own training runs actually use.
Image Classification
Held-out tagging benchmark with per-tag frequency strata and a published error analysis for every baseline.
Image Classification
Drop an image in and read the tags back, with per-tag confidence and the threshold exposed as a slider.
Image Classification
Upload a folder, get it ranked. Mostly built to sanity-check the scorer against your own taste.