To accelerate the study of wildlife populations, Google has open-sourced SpeciesNet, an AI model capable of identifying animal species from camera trap images.
These camera traps, digital devices triggered by infrared sensors, are vital for researchers globally. Though effective, they produce vast quantities of data, demanding a lot of time for analysis, often spanning days or weeks.
Seeking to make this task quicker, Google, through its Google Earth Outreach philanthropic program, launched Wildlife Insights approximately six years ago. This platform allows easy online collaboration, letting researchers share, identify, and analyze wildlife imagery, reducing the time required for camera trap data processing.
Leveraging the power of SpeciesNet, Wildlife Insights offers a suite of analytical tools. Google says this AI model was developed using a massive dataset, including over 65 million publicly available images, as well as contributions from renowned institutions such as the Smithsonian Conservation Biology Institute, the Wildlife Conservation Society, the North Carolina Museum of Natural Sciences, and the Zoological Society of London.
Capable of classifying images with remarkable precision, SpeciesNet can assign one of more than 2,000 distinct labels, spanning individual animal species, broader taxonomic categories like “mammalian” or “Felidae,” and even non-animal objects such as “vehicles.”
In a blog post published on Monday, Google said:
The SpeciesNet AI model release will enable tool developers, academics, and biodiversity-related startups to scale monitoring of biodiversity in natural areas.
Released under the Apache 2.0 license on GitHub, SpeciesNet is widely accessible for commercial use with minimal limitations.