Team 46 - DishIt

Inspiration

In a digital age where many people post pictures of everything in their lives, it's common that many people may want to emulate what they see. DishIt was created to help people replicate the food they see on social media with affordable, nearby resources.

What it does

DishIt uses machine learning and takes an image to classify the food in the image, and then scours the web for recipes that you can replicate.

How we built it

We used Machine Learning with Tensorflow, training our own model with a dataset of 101000 images of food, in order to detect the dish of any images inputted into the website with a 75% accuracy. We then used the Recipe API to get the top 3 recipes of that dish.

Challenges we ran into

The model was rather bothersome to train as we came across many bugs forcing us to restart our training multiple times. This was also our first time using React.JS to build our front end and we came across many difficulties building pleasing UI for the user.

Accomplishments that we're proud

We are proud of the fact that we were able to fully train our own ML model and also successfully used React.JS for the first time.

What we learned

We learned how to use Tensorflow.js to train and call from our own model in order to create a web-based Machine Learning application such as DishIt.

What's next for DishIt

Our next steps are to make the UI cleaner, the model far more competent in identifying food items, and to introduce more settings and options such as calorie count, sugar count, and finding the cheapest option for ingredients near your location. We also plan to display online options and allow the user to select dietary options for recipes including vegetarian, vegan, gluten free, and lactose free options.