NOTE: This is a continuation hack

Inspiration

Hannah suffered from severe allergies when she was a child, causing her family to spend hours at grocery stores verifying that the food products would be safe for her. However, not everyone always has enough time. Over 200,000 people are hospitalized every year due to allergies, and 520 million people around the world suffer from them.

Having a simple and fast way to verify the safety of food products is more imperative than ever during this time of quarantine, as we need to reduce the number of people shopping.

What it does

Aller-Free reads ingredient labels to alert users if they are allergic or not.

  • Instant scans
  • Scans anytime, anywhere
  • Only allergy detection app that can scan without an internet connection (and 88% of 200 people with allergies surveyed said they wanted this feature)
  • Scans for up to 300 derivatives (latest alpha version only)
  • Scans for foreign languages and automatically translates Latin text, allowing users to scan any Latin ingredient, and we have built alpha versions that also support non-Latin languages such as Chinese, which is perfect for travelers.

How we built it

We used Google Firebase's ML Kit to read ingredient labels, Google Cloud Translate to translate detected Latin languages into English to compare with users, and we have built alpha versions using Tesseract OCR and Google Cloud Vision to scan non-Latin languages for true availability anytime, anywhere. All of this is built using Flutter, a cross-platform framework based on Dart.

Challenges we ran into

We ran into many challenge - too many to note here.

Accomplishments that we're proud of

  • We have had nearly 100 downloads in the past two weeks
  • At its peak last week, our product scanned 14,000 ingredients daily

What we learned

  • All about programming apps, building with Flutter, and using APIs

What's next for Aller-Free

  • Enhanced foreign language support, including without internet (by loading our own translation models on-device)
  • Increased accuracy (by developing our own ML models!)
  • Increased safety (by having our own database of ingredient labels loaded on a global network of servers that can fix incorrectly scanned products when users are connected to the internet)
  • Better support for false positives and false negatives
Share this project:

Updates