Detectors trained on your objects, not a generic zoo

Same motion as language and physics: start from a proven YOLO base, adapt it to your classes and cameras, and ship the weights with the labeled set. First public models and datasets land this week.

Talk about a custom detector Hugging Face

Status

Soon

Custom YOLO checkpoints and datasets will appear on this grid — model page plus nested data page, same pattern as Flow.

What will live here

Adapted YOLO

A named checkpoint per domain: classes, input size, and a commercial KPI (mAP, latency) — not a research dump.

Paired dataset

Nested under the model. Splits, license, and a path to a private labeling campaign if your images cannot go public.

Custom campaign

Your cameras, your defect taxonomy, your plant. Research releases stay CC-BY-NC unless we agree otherwise.

Different classes, cameras, or a private set?

Book a technical call