Skip to content

image classification using a convolutional neural network

Notifications You must be signed in to change notification settings

larakollokian/foodalike

Repository files navigation

foodalike

have u ever been called a lobster bisque? do boys dm u about ur baby back ribs? find out which scrumptious food u r!!!

check out our devpost
check out our app on heroku
our android frontend

what it is

Take a selfie from the foodalike app and find out what food you look like ;)

We chose 24 categories from the food-101 dataset and trained a Convolutional Neural Network with two laters over 10 epochs. We achieved an accuracy of ~ 65% and a validation accuracy of ~ 35%. We then set up a python backend tried to set up a python server on heroku with flask to handle http post calls to get predictions from the trained model. Our android app sends the user's selfie as an http requests to the backend and gets the category name as a response. It then outputs a short description

who we are

Lara Kollokian
Marie Vu
Luke Ma
Ethan Itovitch

what's next

We made an improved model and we want to train it on a more powerful computer to get better predictions. We also want to successfully connect the server to the android app. Our backend currently only runs locally :(

Releases

No releases published

Packages

No packages published

Languages