Google launches Thinesnet, an Amnesty International model designed to determine wildlife

Google has Open source Thingsnet, designed to identify animal species by analyzing images from camera traps.

Researchers all over the world use camera traps – digital cameras connected to infrared sensors – to study wildlife groups. But while these traps can provide valuable visions, they generate huge amounts of data that takes days to the removal.

In an attempt to help, I launched Google Wildlife Insights, an initiative for the company’s charitable work program in the company’s Google Earth Outreach, about six years ago. Wildlife Insights provides a platform where researchers can share, identify and analyze wildlife images on the Internet, and to accelerate the camera trap data analysis.

Several Life Life Insights are run by Typesnet, which has been trained by Google on more than 65 million pictures and photos available to the public from organizations such as the Smithsonian Institute of Biology, the Wildlife Preservation Association, the North Carolina Museum of Natural Sciences, and the Animal Association in London.

Google says Typesnet can classify images into one of more than 2000 marks, covering animal species, varieties such as “mammals” or “Felidae”, and non -automatic things (for example “vehicle”).

“The TypesNet AI model will enable the tool developers, academics and startups related to biological diversity to expand the scope of biological diversity in natural regions.” He wrote in a blog post published on Monday.

Typesnet is available on GitHub under APACHE 2.0 license, which means that it can be used to a largely commercial restriction.

It should be noted that Google’s is not the only open source tool to automate camera trap photos. Microsoft’s AI maintains a good laboratory Pytorch wildlifeThe artificial intelligence framework that provides pre -trained models has been seized to detect and classify animals.