Google has released an AI framework called SpeciesNet, aimed at recognizing animal species through the evaluation of images captured by camera traps.
Researchers globally utilize camera traps—digital cameras integrated with infrared sensors—to monitor wildlife populations. While these devices offer crucial insights, they produce vast amounts of data that can take days to weeks to analyze.
To assist with this challenge, Google introduced Wildlife Insights roughly six years ago as part of its Google Earth Outreach philanthropic initiative. This platform allows researchers to collectively upload, identify, and scrutinize wildlife images online, facilitating faster analysis of camera trap data.
Many of the analytical tools provided by Wildlife Insights leverage SpeciesNet, which Google asserts was developed using over 65 million publicly accessible images, alongside contributions from organizations such as the Smithsonian Conservation Biology Institute, the Wildlife Conservation Society, the North Carolina Museum of Natural Sciences, and the Zoological Society of London.
According to Google, SpeciesNet can categorize images into over 2,000 distinct labels, encompassing various animal species, taxa such as “mammal” or “Felidae,” and non-animal entities (e.g., “vehicle”).
“The launch of the SpeciesNet AI model will empower developers, academic researchers, and startups focused on biodiversity to enhance the monitoring of ecosystems,” Google stated in a blog post released on Monday.
SpeciesNet has been made available on GitHub under the Apache 2.0 license, permitting commercial use with minimal restrictions.
It’s important to highlight that Google is not the sole provider of open-source tools for the automation of camera trap image analysis. Microsoft’s AI for Good Lab also offers PyTorch Wildlife, an AI framework that includes pre-trained models specifically designed for animal detection and classification.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
