Wildlife researchers develop field-scouting software to identify wildlife in camera trap images


A key problem for practitioners using camera traps for monitoring wildlife is that vast amounts of manual time is spent reviewing the many thousands of images, collected by camera traps out in the field.

Dr Greg Falzon and a team of software engineers based at the University of New England have developed ClassifyMe, a software tool which allows field ecologists and wildlife managers access to the latest advances in artificial intelligence to automatically filter, sort and identify target species, in the captured images.

The tool is designed to fill a major gap in the use of camera traps for the last 20 or so years. There are still significant hurdles to overcome in using computational methods to automate camera trap image processing, but it is progressing well.

“Training computers to identify the hundreds of thousands of animal species globally across thousands of habitats is challenging,” said Dr Paul Meek of NSW Department of Primary Industries, who helped road-test and develop the software.

“We are not content with models that don’t get it right a high percentage of the time, so we continue to work on algorithms to improve the application of ClassifyMe”.

The ClassifyMe software provides users the opportunity to utilise state-of-the-art image recognition algorithms without the need for specialised computer programming skills. ClassifyMe is especially designed for field researchers, allowing practitioners to process camera trap images using field computers instead of office-based workstations. Although the team has designs for a Cloud based portal that will use similar tools to process camera trap images.

“ClassifyMe is designed to save us thousands of hours manually reviewing images, which means I have more time to put my skills to being out in the field-testing new products and developing solutions to wildlife management issues.

“We wanted to make sure we develop software which can be used freely by all researchers across the globe,” Dr Meek said.

This research was funded by the ‘Wild Dog Alert’ research initiative delivered through the Invasive Animals Cooperative Research Centre (now Centre for Invasive Species Solutions), with major financial and in kind resources provided by the Australian Government Department of Agriculture and NSW Department of Primary Industries, University of New England, Meat and Livestock Australia and Australian Wool Innovation.

The research behind this product was recently publish the journal Animals.

Falzon, G.; Lawson, C.; Cheung, K.-W.; Vernes, K.; Ballard, G.A.; Fleming, P.J.S.; Glen, A.S.; Milne, H.; Mather-Zardain, A.; Meek, P.D. ClassifyMe: A Field-Scouting Software for the Identification of Wildlife in Camera Trap Images. Animals 202010, 58. https://doi.org/10.3390/ani10010058