This is a research benchmark under active development. Only the latest main is
supported; there are no backported fixes.
Please report security issues privately via GitHub Security Advisories rather than a public issue. We will acknowledge and respond as soon as we can.
This is the most important note here. DDI (Daneshjou et al. 2022) is gated via
the Stanford AIMI portal under a Research Use Agreement that forbids
redistributing any portion of the dataset. DDI images, derived fold manifests, and
per-image outputs are gitignored and must never be committed; only aggregate
metric summaries (runs/rq*.json) are tracked. Regenerate DDI artifacts locally
after accepting the agreement. HAM10000 is open and its frozen split manifests are
included.
Everything runs locally and sends nothing to external services. This is a research benchmark for studying evaluation methodology. It is not a medical device and must not be used for diagnosis or patient care.