Fastai is a library that sits on top of pytorch and essentially makes training machine learning models easier. I’ve learned how to make basic classification models (including multi-label classification), perform image segmentation on the CamVid dataset, work with keypoint datasets, and deploy a model to production. Also, I have gained an understanding about how to do collaborative filtering and machine learning with tabular data. Specifically, I deployed a fruit classification model trained on the Fruits-360 dataset to Heroku using node.js with python as a backend. I also used flask to deploy the same model. The course has helped to learn more about machine learning in general. For example, I now understand SGD and how it is used during training.
Stanford CS231n is teaching me more about the theory of Computer Vision, with a specific focus on Convolutional Neural Networks.
Project link: https://github.com/HHousen/fruit-classifier-app-node