Google launches an artificial intelligence model to identify wildlife species.

Google has launched an artificial intelligence tool designed to identify animal species, aimed at facilitating wildlife monitoring. This new tool, called SpeciesNet, was announced on Monday and is presented as an open-source model intended for wildlife biologists.

SpeciesNet includes a model that allows for the identification of objects in recordings from cameras that monitor wildlife, as well as another model that classifies those objects into different animal species. Since 2019, wildlife biologists have had access to SpeciesNet through Wildlife Insights, a cloud-based tool from Google. Now, it has been released to the public in open-source form.

Scientists often use motion-activated cameras to study animals in their habitats, but processing the recorded material can be slow, as it involves reviewing vast amounts of images. As mentioned in the SpeciesNet repository on GitHub, "AI can speed up this processing, allowing conservation professionals to spend more time on preservation and less time reviewing images."

Google reports that SpeciesNet was trained on a dataset that includes over 65 million images, covering both images from camera traps submitted by users of Wildlife Insights and publicly available datasets. This model combines information gathered from its underlying models to make predictions about each identified animal and displays a percentage of accuracy.

With the ability to classify images into more than 2000 labels, SpeciesNet covers various animal species, higher taxonomic groups (such as 'mammalia' or 'felidae'), and non-animal classes (such as 'empty' or 'vehicle'). This tool is currently available as an open-source model on GitHub.