Capture with care.
Choose a clear, well-lit photograph. Center one skin spot and keep the area in focus.
A MOMENT OF ATTENTIONA NEW PERSPECTIVE ON SKIN AWARENESS
A closer look.
A clearer
perspective.
Skin deserves your attention.
A private space to explore a skin concern and prepare for a more informed conversation.
Research demo · For exploration, not diagnosis
OBSERVE. UNDERSTAND. TAKE CARE.
THOUGHTFULLY DESIGNED. PRIVATELY PROCESSED.Discover the approach
01 / THE APPROACH
A little attention.
A meaningful first step.
From a photograph to a little more context. A simple experience, with its possibilities and limitations made clear.
Explore the pattern.
An experimental model compares your image with patterns learned from research images.
CONTEXT, WITH LIMITSStart a conversation.
Review the explanation and bring any concerns to a qualified healthcare professional.
CARE BEYOND THE SCREEN02 / EXPLORE THE DEMO
Your skin.
Your space.
Take a moment to look closer. On-device analysis. Optional photo sharing with Groq.
For a clearer photograph
- Use bright, even lighting without glare.
- Show one spot clearly; the entire image is resized for analysis.
- Keep the camera steady and the spot in focus.
This research demo cannot diagnose a condition or rule out cancer. It can also produce a result for unsuitable images. Seek professional advice about any new, changing, bleeding, or concerning spot.
PRIVATE PHOTO ASSESSMENTEXPERIMENTAL
A closer look, in progress.
Loading the research model…
Your photograph stays in this browser.YOUR PHOTO / RESEARCH REVIEW
A clearer overview.
Your assessment, explained.
LOCAL MODEL ASSESSMENT
The model cannot identify a specific disease. A lower signal does not rule out cancer; a higher signal is not a diagnosis.
03 / PRIVATE BY DESIGN
Some things
should stay
with you.
Your skin photographs are personal. On-device analysis. Optional photo sharing with Groq.
Explore with privacy04 / AN HONEST PERSPECTIVE
Clarity starts
with transparency.
This is an early research project. We believe understanding what a tool cannot do is just as valuable as exploring what it can.
20/29
Melanoma examples flagged for review
In the saved research test, the review threshold flagged 20 of 29 melanoma images and missed 9. Across all cancer types it flagged 237 of 257 cancer-labeled images, but also 192 of 256 benign-labeled images.
Flagged includes uncertain results. At this threshold, binary research accuracy was 58.7%, cancer sensitivity 92.2%, and specificity 25.0%. These are offline research results, not clinical validation or an individual cancer probability. Browser resizing may produce different scores. The model weights have not been retrained.
A few things to know.
Can Derma Scope tell me if I have skin cancer?
No. It compares image patterns with two research categories. It cannot diagnose cancer, identify a specific condition, or replace an examination by a healthcare professional.
What happens to my photograph?
On-device analysis. Optional photo sharing with Groq. Selecting another photo or refreshing the page clears it from the active experience, but cannot retract a photo already shared with Groq. Provider retention rules apply.
Why isn’t there a confidence percentage for my photo?
The model's output scores have not been clinically calibrated. Displaying them as a personal cancer probability would be misleading. We show the experimental category, its limitations, and guidance about professional care.