Deep Learning Food Recognition Model

Artificial Intelligence Fundamentals Course Project 2019

University of Toronto

APS360: Artificial Intelligence Fundamentals Course Project 2019

Instructor: Prof. Lisa Zhang

Github repo

We, as a team of three, aimed to create a module that would identify food categoreis from an overview photo of a meal. We trained a supervised machine learning model that can perform detection from unseen images quickly and conveniently, then propose bounding boxes for food seen in the image, finally classify food categories from the input image of a meal.

The machine learning model we used was Faster R-CNN (ResNet based).

Faster R-CNN Model Architecture
The processed data has the resided image, ground truth bounding boxes, and labels.
Sample outputs from our model.