markstanl/math443-image-learning-models

Math 443 Final Project Models

This project contains image learning models for group 4's final project for UW-Madison's Math 443: Applied Linear Algebra. There are two main directories in this project, oxford-pets and cityscapes. Each repository contains the models trained on those datasets respectively.

Oxford Pets

oxford-pets contains two models for the dataset, the image classifier and image masker. The ensemble model of these two is roughly a single-object segmentation model. These were trained from random weights, and gave decent results.

CityScapes

cityscapes contains a single model for making instance segmentation predictions on the CityScapes images. The model was fine-tuned from the torchvision Mask R-CNN w/ ResNet50 backbone.