Entomology, Department of
Date of this Version
2022
Document Type
Article
Citation
Master of Science Degree Project, Department of Entomology, University of Nebraska-Lincoln, 2022.
Abstract
Photo-based identification apps for mobile devices have grown in popularity across many fields of organismal biology. While studies have shown that these apps show promise in fields such as botany and ornithology, there is much skepticism surrounding their accuracy and usefulness in entomology. While providing these apps with photos of high quality will undoubtedly lead to the highest accuracy rates, often users submit photos of average quality, e.g., in poor lighting, out of focus, from far away, etc. To quantify realistic accuracy rates of photo-based insect identification apps, this study analyzed three popular apps – Seek, Picture Insect, and Google Lens – using photos of variable quality to simulate realistic/typical use. Taxonomic specificity of app results was also quantified and used to determine relative usefulness of each app. Seek produced the highest accuracy rates of the three, but often yielded taxonomically obvious results, resulting in low usefulness rates across all photo quality tiers. Picture Insect and Google Lens produced moderately high accuracy and usefulness rates for photos of ideal quality, but ≤54% accuracy for photos of acceptable and poor quality.
Comments
Copyright © 2022 Morgan Manderfield