Food.ai

• 

  UI

• 

  Food Detection

Inspiration

Many people nowadays are health conscious and want to keep track of their daily calorie count and we made an app to keep track of that using an efficient object detection machine learning algorithm combined with a simple UI to provide data about daily calorie intake.

What it does

Seamlessly detects foods and tracks your daily calorie count based on what you ate.

How we built it

We applied a technique called transfer learning to train our custom deep learning model on food datasets.

Challenges we ran into

Finding good food datasets were not as easy as we initially thought.

Accomplishments that we are proud of

We successfully created a working object detection app using deep learning algorithms, as well as a UI in Android Studio that utilizes this tensorflow model in order to obtain calorie information about the detected foods, helping users keep track of their daily calorie intake.

What we learned

In addition, Tensorflow is often deprecated and models take a long time to train, even when using transfer learning. Homogeneity in the dataset was also an issue (not all pizzas are triangle-shaped :))

What's next for Food.ai

We plan to train on a larger variety of foods with a larger, more robust dataset as well as add location services to the app so restaurants can input calorie information for foods that differ significantly from their normal calorie count.

Built With

• android-studio
• java
• numpy
• python
• tensorflow