Google has open-sourced the code of its AI model called SpeciesNet , which can identify animal species by analyzing photos from camera traps. These cameras, used by researchers worldwide, provide valuable information about wildlife populations but generate vast amounts of data that can take days or even weeks to process.
To streamline this process, Google launched the Wildlife Insights project about six years ago. This initiative allows researchers to share, identify, and analyze wildlife images on an online platform, accelerating the processing of camera trap data. Many of the analysis tools in Wildlife Insights use SpeciesNet, which was trained on more than 65 million publicly available images, as well as materials from organizations such as the Smithsonian Conservation Biology Institute, Wildlife Conservation Society, and others.
SpeciesNet can classify images into more than two thousand categories, including animal species, taxa such as “mammals,” as well as non-animal objects like “vehicle.” Google notes that releasing the SpeciesNet model will enable tool developers, scientists, and biodiversity startups to scale biodiversity monitoring in natural areas.
SpeciesNet is available on GitHub under the Apache 2.0 license, allowing commercial use with minimal restrictions. It is worth noting that Google is not the only company offering open tools for automating camera trap image analysis. For example, Microsoft AI for Good Lab supports PyTorch Wildlife, an AI framework that offers pre-trained models for animal detection and classification.
