A smartphone-enabled progressive web application that supports machine learning education

Teaching core concepts in artificial intelligence (AI), such as supervised machine learning (ML), often involves computer coding exercises. Prior experience in coding and familiarity with specific command-line or graphical interfaces often impact the student’s learning experience and educational attainment in this area. Also, provisioning a suitable computer system that allows students to run these coding exercises can pose significant challenges to teachers, especially in the K-12 setting or some interdisciplinary settings such as medical education. Inspired by the concept of edge computing, which focuses on using the computing resources within the student’s digital devices such as smartphones or tablets, we have developed a progressive web application (app) that can perform ML training and testing using camera-taken photos as input data. As an initial example, we have developed an app that allows the training of a deep convolutional neural network (CNN) on the student’s smartphone such that the phone can detect pneumonia from chest X-ray (CXR) images taken from the device’s camera. The app contains a pre-trained CNN on an extensive collection of labelled CXR images, which greatly reduces the number of training samples a user needs for training.

The data preprocessing pipeline and CNN are implemented in JavaScript and deployed into the app. A user-friendly web interface is constructed through a front-end framework to maximise usability. We tested this app in several introductory ML workshops for medical students and high school students. These workshops aimed to let the students experience the process of collecting and labelling training data (prepared CXR images) and train their CNN-based pneumonia detectors on their smartphones. With this app and corresponding dataset, students could learn ML anywhere and anytime without coding experience or the need for centralised computing resources. We believe our effort is an important step in making AI education more accessible.