About

This repository utilizes PyTorch framework in order to classify food images in a in-class Kaggle challange. Repository is a proposed solution as part of an Applied Machine Learning course at University of Amsterdam MS Information Studies: Data Science.

Description

1. Dataset_EDA.ipynb - Visualizes provided dataset in order ot inspect for any data imbalance problems.
2. Dataset_Modifications.ipynb - Implements data augmentation techniques to balance class numbers.
3. Ensemble_PyTorch_Model.ipynb - Implementation of simple commity of networks model ensembling.
4. Kaggle_Test_Predictions.ipynb - Script which performs class prediction and outputs in nesessary Kaggle format.
5. Lasagna_GAN.py - Initial implementation to battle class imbalance using GANs.
6. Loss_Analysis.ipynb - Visualization of saved history objects in order to analyze loss vs accuracy dueing training epohs.