This project aims to create a neural network based classifier that can successfully classify certain seafood type species. The network architecture will maintain a basic structure consisting of a couple of convolutional and pool layers followed by fully connected layers at the end. The dataset that will be used for the project consists of a total of 430 images of 9 species of fish types [1]. The images have three channels and resolution of either 2832 x 2128 or 1024 x 768. The project uses the Kornia framework in an attempt to improve generalization by introducing different data augmentations prior to model training.
Note that the project is based on the cookiecutter data science project template.