This is a simple temporary app that was used to deploy a ResNet-50 model on Heroku cloud. The model was trained on Food-101 dataset from Kaggle.
The jupyter notebook in this repository contains all the steps I've taken to train the ResNet on this dataset. The model was only trained for 5 epochs and acheived an accuracy of 76% on test dataset. Further improvement of the accuracy can be made in future releases, but for now I decided to go with this temporarily to set up the DevOps pipeline.
The app is available on Heroku. It's likely that you may not find the app to be working as Heroku has become a paid platform and I probably ran out of money. 😆