This repository contains the code and resources used for our project on face image-based emotion detection using neural networks. This project was conducted as part of the "Artificial Intelligence and Information Management" lecture, culminating in a Kaggle competition where we secured second place with a weighted F1-score of approximately 0.855.

• notebook.ipynb: This notebook contains the method that achieved our best result in the competition. It includes the final model and the data preprocessing and augmentation techniques used.
• scratch-model.ipynb: This notebook shows our custom-designed Convolutional Neural Network (CNN) built from scratch. It details the architecture and training process of our initial model.
• data-analysis.ipynb: This notebook focuses on the analysis and augmentation of the data. It includes exploratory data analysis (EDA) and the various data augmentation techniques applied to improve model performance.
• oversampling.ipynb: This notebook demonstrates our approach to handling class imbalance through oversampling and the use of stratified splits. It details the methods and their impact on model performance.
• ensemble.ipynb: This notebook explores future improvements of the model using ensemble learning techniques. It includes preliminary implementations and discusses potential benefits.