The purpose of this notebook is to classify images of food into 101 categories.

The paper we will try to follow is Food-101 – Mining Discriminative Components with Random Forests

This notebook was inspired by Daniel Bourke's version

1. We will be developing our model in TensorFlow.
2. The first layer will be a pretrained EfficientNet with all the layers frozen.
3. This is followed by a softmax with 101 classes.
4. After training the model we will unfreeze the last 5 layers of the EfficientNet layer and train it again.

• Food-101 – Mining Discriminative Components with Random Forests
• Daniel Bourke's version