Training & fine-tuning
Train or fine-tune classification and foundation models on real clinical cases with one diagnosis per patient, from smartphone and professional-camera sources.
Partnered with DermNet NZ
De-identified, dermatologist-diagnosed skin images, licensed for training, validating, and deploying medical AI.
Complimentary evaluation samples · Custom selections · Quote on request
Our collections come from in-person dermatology consultations across independent practices and countries. All are private and licensed directly by CoUnity Data, and we assemble custom selections to fit your model.
Main collection
Each case is one patient consultation, with one primary diagnosis made in person by the treating dermatologist.
Real-world smartphone photographs plus professional-camera imagery, to train and evaluate across capture conditions.
Cases from multiple countries and practices, spanning a range of skin tones.
Subsets by condition, volume, demographics, or capture type, with complimentary evaluation samples.
Condition-level counts and distributions are shared during scoping.
Every case links clinical photographs to a single diagnosis and a small, consistent set of fields, so you can split, stratify, and audit your data with confidence.
Fields shown are for the Main dataset. Exact field names, and any additional fields in other collections, are documented with each sample.
Data is collected in clinical practice and goes through manual review before it is offered for licensing.
Photographs are taken during the visit, and the treating dermatologist records the diagnosis.
Collection is designed around one case per unique patient, with residual duplication estimated below 1%.
Each case is checked for duplicates, image quality, age and sex accuracy, and de-identification.
Records carry de-identified IDs, not names or contact details, and are licensed only under a signed agreement.
HIPAA-aligned de-identification. Our documented patient de-identification process is aligned with HIPAA Safe Harbor methods. We are happy to walk your compliance team through it during scoping.
A short, predictable process with one standard agreement. Pricing is volume-based and quoted per project.
Book a call or email us with the conditions, volume, capture type, and intended use you have in mind.
We share a complimentary evaluation sample so your team can check image quality and metadata.
We propose a selection from 1,000 images upward. Pricing depends on volume, conditions, and independent-case coverage.
Review and sign our standard commercial licensing agreement with CO-UNITY GROUP LLC.
A ZIP of JPEG images and a JSON manifest, typically within 72 hours of signature and payment.
Train or fine-tune classification and foundation models on real clinical cases with one diagnosis per patient, from smartphone and professional-camera sources.
Build held-out test sets from independent practices and capture conditions to measure how models perform on real-world images before release.
Ship products built on licensed data. The perpetual commercial license covers deployment and derived applications, and you own the models you build.
Share the conditions, volume, and intended use you have in mind. We'll come back with a sample and a quote for a custom selection.