Artificial Intelligence's Role in Global Camera Trap Data Management and Analytics via Wildlife Insights

By

Thau, David
Ahumada, Jorge A., 1965-
Birch, Tanya
Fegraus, Eric
Flores, Nicole
Jetz, Walter
Kays, Roland
Kinnaird, Margaret
Kulkarni, Sayali
Lyet, Arnaud
O’brien, Timothy G.
Palmer, Jonathan
Schuttler, Stephanie
Duong, Michelle
Oliver, Ruth
Zhao, Jennifer Y.
Mcshea, William J

Type

Article

Date of Publication

2019-7-17

Original Publication

Metabarcoding and metagenomics

Volume

3

Pages

1--2

Subjects

Artificial intelligence , Big Data , computer vision , species identification

Contributed by

Pensoft Publishers

Language

English

Published by

Bulgaria Pensoft Publishers 2017-

Identifiers

DOI: https://doi.org/10.3897/biss.3.38233

 

Abstract

Camera traps have existed since the 1890s (Kucera and Barrett 2011), but they weren’t widely used until the introduction of commercial infrared-triggered cameras in the early 1990s (Meek et al. 2014). Since then, millions, perhaps billions of camera trap images have been collected for many reasons, biodiversity monitoring being one of the key applications. Unfortunately, although there are camera trap deployments all over the world, these operations occur in isolation, limiting the impact they could have on a global understanding of biodiversity health. Even within individual institutions, managing and analyzing multiple camera trap deployments in aggregate can be challenging. In fact, managing a single deployment of camera traps is non-trivial and important data are frequently cast aside as bycatch, left unanalyzed on decaying hard drives.<br> Wildlife Insights attempts to overcome these hurdles by providing camera trap data upload, management, and analysis services. It provides the world’s largest database of camera trap images by bringing together the camera trapping efforts of several the world’s largest conservation and research organizations, and it is open to future contributors. Artificial Intelligence-driven services sit at the heart of the platform. New camera trap data uploads are automatically analyzed to differentiate between images with people, non-human animals, and no animals. The images with non-human animals are further analyzed to detect specific species. The proposed labels are sent back to the submitter for review and then uploaded to the database. All uploaded images, unless specifically embargoed, are immediately available for analysis by all users of the system. A selection of tools are provided to support analyses of global biodiversity.<br> This presentation will describe Wildlife Insights and its AI implementation in detail, contextualized by case studies using analyses of the data currently stored on the platform. Challenges around integrating camera trap data within the platform and with other external services that work with the platform will also be discussed. The talk will end with some thoughts about future directions for the AI services, especially with regards to integration with related platforms.