Australian Wildlife Conservancy Granted Federal Funding For AI Wildlife Identification

The Australian government will inject $750k into Australian Wildlife Conservancy’s cutting-edge AI wildlife-recognition system to help accurately identify 120 species of mammals and reptiles.

 

At present, AI-enhanced camera traps used by the Australian Wildlife Conservancy (AWC) can distinguish 44 species with over 95% accuracy. In 2020, these cameras processed 4.52 million images, and the program can differentiate ‘false positives’, such as moving vegetation, from the genuine presence of animals.

Following further development and training of the AI models, the system will be able to identify 120 species, including threatened species such as the Northern bettong, Western quoll, and the Great Desert skink, showcasing improvements in image processing speed, accuracy, and cost-effectiveness.

‘AWC is increasingly turning to emerging technologies to improve efficiency and the quality of data collected in the field’, says Tim Allard, AWC CEO.

‘This project builds on AWC’s existing sanctuary and data collection infrastructure, bolstered by an established network of expert ecologists familiar with the different fauna assemblages of each region. These resources cannot be replicated without millions of dollars of investment and years of effective project planning’, he said.

How do AI ‘wildlife recognisers’ work?

Just like facial recognition tools, the AI-enhanced program will identify the physical characteristics of a species and match them to data collected from years of biodiversity research around Australia. It uses existing camera trap technology (which has revolutionised wildlife monitoring over the years) bolstered by AI technologies to process the data.

This will save ecologists possible months of manual data processing so that they can spend more time in the field where they have the biggest impact.

Raquel Parker, a former AWC Ecologist who, like other ecologists, relies heavily on remote camera traps to monitor wildlife, appreciates the efficiencies of AI technology.

‘Machine Learning has sped up the process of sorting through large data sets, which means we can spend more time in the field’, Parker said.

The Species Classifier model will be made available to the public with guidelines on how to use the technology to identify animals. AWC will also document and share its camera trap processing methods and tools.

AI has the potential to improve the speed of camera trap image processing

However, the widespread application of AI has its challenges, as a large number of high-quality images are needed of each species in varying habitats in order to train the AI technology accurately. In anticipation of the new technology, the AWC’s National Science Team has started collating images from the organisation’s vast library of photos gathered over 15 years from sanctuaries and project partners.

More information on the project and other conservation works can be found on the Australian Wildlife Conservancy’s website.

Feature image thanks to the AWC