foodscan

• 

  Apple detection w/ 80% prediction accuracy.

• 

  Corn on the cob detection w/ 50% accuracy.

• 

  Detection page. Records any detection from the camera app, and records information to a temporary database.

• 

  Sign-in page. Ingests user log-in from other third party authenticators to store data, caches, and more inside the app.

Inspiration

Every day, I usually spend about 10 minutes trying to find a recipe or meal to eat with the food at home. Something > healthy, that I can mix together into a meal with some healthy green stuff too :)

This is why I created FoodScan+.

What it does.

The app has datasets over 100,000 photo assets of food that have been run through a machine learning program. Once the camera is activated, it will sense for the same image as learned from before. As in my screenshots, I have > gotten a certain possibility (in %), as to what my fruit/vegetable would be. This app also can create recipes off the > grocery that you have scanned, with lots of filters to become as specific as you want!

How I built it!

Using previous machine learning applications, I have gathered enough information to possibly update to a better,
and more advanced dataset. With this decision in mind, I have planned to move towards Google Cloud's AutoML > interface with Google Cloud databases, to store datasets and run them 24/7.

Challenges I ran into...

Errors in runtime, package corruption, and data corruption were the top challenges going through developing and > maintaining this hack. It was definitely more difficult since I have not done something this advanced, but it seemed > like a great challenge.

Accomplishments that I'm proud of!

Made an ML App on Android.. so that I can possibly update more and more in the future. This app is really a good > side-project :) I also learned more Python, went through a different software(s), and implemented APIs I haven't seen > or touched before.

What I learned.

I learned a lot! As said before, it's more of the exploration that usually guides and helps your hack grow, as I never
wanted to do something I knew. That's the point of hackathons in my opinion :)

What's next for FoodScan+

Larger amounts of ML data, more streamlined user-experience, wider range of supporting devices (web-app/swift). I have lots of plans going for this app in the future!

Built With

• c++
• gcp
• google-colab
• java
• machine-learning
• python