Google's SpeciesNet AI Helps Conservationists Identify Animals in Camera Trap Photos

Google's open-source AI model SpeciesNet is helping researchers around the world analyze millions of camera trap photos to protect wildlife—turning a time-consuming task into something that takes days instead of decades.

SpeciesNet, which Google open-sourced about a year ago, can automatically identify nearly 2,500 categories of mammals, birds, and reptiles. The model has been available since 2019 via Wildlife Insights, but the open-source release has expanded access significantly.

In Tanzania's Serengeti National Park, the Snapshot Serengeti project used SpeciesNet to analyze a backlog of 11 million photos—"processing decades' worth of data in just days," according to Google. The project, run in collaboration with the Tanzanian Wildlife Research Institute, previously relied on online volunteers but couldn't keep up with the image volume.

Colombia's Humboldt Institute uses SpeciesNet through Wildlife Insights to monitor species in the Amazon rainforest. The institute recently launched Red Otus, a national camera trap network that has analyzed tens of thousands of images to track changes in wildlife behavior—finding that some mammals are becoming more nocturnal, possibly to avoid human threats.

In the U.S., Idaho's Department of Fish and Game uses SpeciesNet to sort millions of camera trap images each year before human experts conduct final reviews. "Having SpeciesNet sort the images by species beforehand greatly speeds up reviewing," Google noted.

In Australia, the Wildlife Observatory trained SpeciesNet on local species not in the original model to monitor iconic animals like cassowaries.