By combining existing cosmetics research, demography and machine learning, Skin.ai is able to analyze patterns across different skin types, ethnicities and ages to provide you with instant feedback on whether the ingredients in the product scanned may cause acne. No product will be 100% guaranteed against breakouts, but when a product does causes you to breakout, your feedback will allow Skin.ai to learn new patterns and discover new potential comodeginic ingredients, helping everyone out in the process.
Unlike other cosmetic apps, our results are statistically sound and verified by experts.
Sign up. Start scanning! It's that easy.
Our app is designed to get better over time! Each user is like an experiment we can use to learn more about acne breakouts.
After scanning a product, get your results back as Low, Medium or High acne potential!
You can scan a barcode, take a picture of ingredients or simply read the ingredients to get acne scores.
We will never sell your data, or share it without your permission!
How to Register
Voice Search for Product
These videos are live screen caps of actual users using Skin.AI. Click on the the videos to learn more!
Ai.skin uses 3 ways to get ingredient informaiton. We can use a barcode scanner to read the product, use Optical Character Recognition (OCR) and a bounding box on an image you take, or Voice Recognition as you read out ingredients. All other apps on the market are bar code only, relying on someone else to register the product and maintain the database. We offer multiple ways to get your information to the system. Once there, our machine learning algorithm will score the likelyhood of acne breakout based on research on comodegenic (clogging) chemicals.
Sorry about that! Try a laptop browser until we can fix that bug. The rest of the site should be fine though.
Acne is a VERY complicated issue for some people. Your diet, hormone levels, genes and activites all play into whether you'll get acne or not. The last piece of the puzzle is whether the ingredients in your cosmetics and toiletries are making it worse. Our model weighs it's scores based on your age, ethnicity and skin type. What might be a perfectly fine product for a 40 year adult with normal skin, would be potentially a big problem for a teen with oily skin. We can't guarantee perfect results, but the exciting thing is that if a product does cause you problems, giving your feedback allows our models to get smarter, even to the point where we can identify new comodeginic ingredients. It's an opportunity to turn bad purchase into research helping everyone!
Your data is only used in generating and improving our acne Machine Learning (ML) model. In the future we may allow advertisers to make product suggestions if you are scanning a particular type of product, but we'll never give them access to your personal information.
EWG.org has compiled a public database full of information on ingredients and their potential effect, but never any information on acne. We want to give our users the best experience possible, so we also include additional information we gathered from EWG.
There has been some research done over the years on comodeginc ingredients, but not alot. Several websites such as Skinsalvation.org have compiled this information as generic scores from 1-5 on the how comodegenic an ingredient is and made the information public online. Using a database of products found in EWG.org, we applied these comodegenic scores to the ingredients in each product. If a product had a total sum of comodeginic scores of 0, it's a 'Unlikely', if it totals 1-2, it's a 'Low' and if it's above 3 it's a 'High'. We then made a million fake users and gave them the demographics of typical Americans. Each person has either a skin type of oil, dry or normal (more research is needed to get the split on skin type). We then assigned at random various products from our database and acne scores. Our model used these fake profiles to generate the first model
Well, everyone has to start somewhere. Worst case scenario, our product isn't as accurate as it could be if we had REAL user data....but real users wouldn't use the app if it didn't help them out at all. So to solve this chicken and egg problem, we generated fake users first. As people start using the app, we'll be retraining our models to work on the REAL data and getting even better results.
Sorry to hear that! But the exciting thing is that if you take the time to send that feedback to our system, we can integrate it into our prediction model and learn from the mistake. Remember, if you hadn't used the app and bought the product, you'd be in the same place, but this way, you can help others (including yourself) make fighting acne easier in the future! It's a set-back, but great for the future!
To learn more, we've included another webpage with our more technical documentaiton for fellow Data Scientists and researchers. To help out scroll to the bottom of this page and contact us!
I knew nothing about cosmetics and acne until I got this app...for that matter I also didn't know crap about html or css. I'm learning so much about what causes acne and will think twice about all my purchases.
All, I just wanted a way to easily scan cosmetic ingredients! Who knew OCR on a phone could be so tough...
It's an impressive little app, considering where we were when we started....you should see my backend!