SCIN: A new resource for representative dermatology images
90% of 16k+ crowdsourced skin images. Google just released SCIN—the first dermatology dataset that actually represents all skin tones.
Why it matters
Google's new SCIN dataset addresses a critical AI blind spot: existing dermatology datasets skew heavily toward lighter skin tones, creating AI tools that fail for darker skin types. This release signals how bias in medical AI training data directly impacts healthcare equity and model performance across demographics.
The key facts
8 to knowSCIN dataset contains 16,000+ dermatology images with balanced Fitzpatrick skin type distribution (Types 3-6 overrepresented vs. clinical datasets)
97.5% of crowdsourced contributions were genuine skin condition images (low spam rate)
~90% of 8-month study contributions released after filtering
80% of contributions include self-reported symptom data (texture, duration, symptoms)
Dermatologist label confidence depended more on self-reported data availability than image quality
Dataset focuses on everyday conditions (rashes, allergies, infections) vs. clinical datasets focused on neoplasms
Open-access resource released via collaboration between Google Research and Stanford Medicine
Crowdsourced via web search result page advertisements to reach people at earlier health journey stages
Go to the source
Google Research Blogblog.research.google
Publisher excerpt: Posted by Pooja Rao, Research Scientist, Google Research Health datasets play a crucial role in research and medical education, but it can be challenging to create a dataset that represents the real world. For example, dermatology conditions are diverse in their appearance and severity and manifest…
