Final Project: Semi-Intelligent Image Classification App | Interactive Art and Computational Design, Spring 2016

Final Project: Semi-Intelligent Image Classification App

The Semi-Intelligent Image Classification App makes it easy for users to train and test their own image classifiers. Let’s say you have a bunch of images that you’d like to categorize based on a simple binary classification. Just select a classifier and throw in the set of images, and the app will classify those images into pre-defined categories, such as Day/Night, or Apple/Orange. You can use one of the Featured Classifiers that I have trained or train your own classifier using the web interface.

The most challenging part of my project was bridging the gulf between the back-end machine learning techniques and front-end user interface. I had to build a powerful, effective tool yet make it easy enough to use for a person who had no prior knowledge about machine learning. For this reason, I added some constraints: I only offered one algorithm (SVM) and restricted the app to binary classifications. So, the user is placed a level of abstraction away from the complexity of the learning and classification that happens in the backend.

Ever since I started learning about machine learning, I’ve always wondered: how can I take this awesome power of machine intelligence and deliver it to the common man? Right now, machine learning is a tool of the technical elite, those who have the knowledge and programming expertise to reap the benefits of AI. My project is a proof-of-concept that machine learning can be made accessible to everyone (with some basic conceptual understanding). From working on this project, I’ve realized that the fundamental goal of developing technology is to solve human problems. And a key part of this goal is making advanced technologies available to the masses in the first place.