Google releases Sinksnet, AI model designed to identify wildlife

Google has Open AI model, Sinkernet designed to identify animal species by analyzing camera traps photos.

Researchers around the world use camera traps – digital cameras related to infrared sensors – to study wildlife populations. But while these traps can provide valuable information, they generate massive volumes of data that take days to weeks to sift.

In an attempt to help, Google launched Insights Wildlife, an initiative of Google Earth’s Philanthropy program on Google Earth, about six years ago. WildLife Insights provides a platform in which researchers can share, identify and analyze wildlife images online by collaborating to accelerate the camera trap analysis.

Many of the wildlife analysis tools are powered by Sinksnet, which Google claims to have been trained in over 65 million publicly available images and images by organizations such as the Institute for Biology of Smithsonian Conservation, the Society for the Protection of the Wildlife.

Google says species can classify images in one of more than 2000 labels covering animal species, taxa such as “mammals” or “felidae” and objects that are not residents (such as “vehicle”).

“The edition of the AI ​​Sixingnet model will enable tool developers, academics and start -ups related to biodiversity to scales biodiversity monitoring in natural areas,” Google wrote in a blog post posted on MondayS

Sinksnet is available to GitHub under the Apache 2.0 license, which means that it can be used to a large extent SANS restrictions.

It is worth noting that Google is not the only opening tool for automation code for the camera trap image analysis. AI for a good Laboratory of Microsoft supports Pytorch WildlifeAI frame, which offers pre -trained models, refined to detect and classify animals.