In this project, I have trained a Convolutional Neural Network (CNN) in PyTorch to classify facial expressions. The project involves loading a pretrained state-of-the-art CNN model and fine-tuning it on a dataset of 48 x 48 pixel grayscale images of faces. The dataset includes seven targets: angry, disgust, fear, happy, sad, surprise, and neutral expressions.
To improve the model's performance, I have applied image augmentation techniques specifically designed for classification tasks. Additionally, I have created train and evaluator functions to facilitate the training loop.