Authoring Platform for Mobile Citizen Science Apps with Client-side ML

TL;DR

This paper presents an authoring platform for creating mobile citizen science apps that incorporate client-side machine learning to improve data collection accuracy and efficiency, demonstrated through biodiversity and rip current detection use cases.

Contribution

The paper introduces a novel platform enabling easy development of citizen science apps with integrated client-side ML guidance, enhancing data quality and collection efficiency.

Findings

Apps assist participants in recognizing correct data.

Increased data collection accuracy and efficiency.

Successful demonstration with biodiversity and rip current detection apps.

Abstract

Data collection is an integral part of any citizen science project. Given the wide variety of projects, some level of expertise or, alternatively, some guidance for novice participants can greatly improve the quality of the collected data. A significant portion of citizen science projects depends on visual data, where photos or videos of different subjects are needed. Often these visual data are collected from all over the world, including remote locations. In this article, we introduce an authoring platform for easily creating mobile apps for citizen science projects that are empowered with client-side machine learning (ML) guidance. The apps created with our platform can help participants recognize the correct data and increase the efficiency of the data collection process. We demonstrate the application of our proposed platform with two use cases: a rip current detection app for a planned pilot study and a detection app for biodiversity-related projects.