not-chilaquiles

An app to detect if the food in your plate is a torta, a burger, a sandwich, chilaquiles, or something else completely. It uses a Convolutional Neural Network called VGG pretrained with the ImageNet dataset, and then retrained and finetuned with the dataset provided to identify 4 categories of dishes. The ConvNet was built on top of Keras with a Theano backend. Then built a minimal Flask server to load the trained model and receive image requests and then reply with the categorical choice and the certainty percentage. The client for this server is an iOS app developed in Swift.

A challenge by Definity First.

Built With

Share this project:

Updates