Your Skin Cancer Model Fails On Real Photos
This story is from 2026-10-26. It is preserved in the archive; the latest stories are on the live feed.
A 10-year-old CNN can beat frontier LLMs at skin cancer diagnosis, but only when the input looks like the training data. In practice, that almost never happens. Your model sees dermoscopic images: high-resolution, ring-light illumination, no hair or text. Your users send phone photos taken under ba…
Read the full story at DEV Community — Machine Learning ↗
Timeline · 1 report
- 2026-10-26 01:34 · DEV Community — Machine Learning
Your Skin Cancer Model Fails On Real Photos