Google Speciesnet: AI Wildlife Identification

Google's SpeciesNet: AI for Wildlife Identification

Google has made its artificial intelligence model, SpeciesNet, publicly available. This AI is specifically engineered to identify animal species through the analysis of images captured by camera traps.

The Challenge of Camera Trap Data

Camera traps – consisting of digital cameras linked to infrared sensors – are widely utilized by researchers globally for the study of wildlife populations. However, the sheer volume of data generated by these devices often requires significant time, sometimes weeks, to process effectively.

Analyzing this data can be a bottleneck in wildlife research. The need for efficient processing methods is critical for timely insights.

Introducing Wildlife Insights

Approximately six years ago, Google addressed this challenge with the launch of Wildlife Insights. This initiative, stemming from Google Earth Outreach’s philanthropic efforts, provides a platform for researchers to collaborate on wildlife image analysis.

Through Wildlife Insights, researchers can share images, identify species, and collectively accelerate the analysis of camera trap data.

SpeciesNet: The Engine Behind the Analysis

Many of the analytical capabilities within Wildlife Insights are powered by SpeciesNet. Google reports that the model was trained using a substantial dataset exceeding 65 million images.

This training data includes publicly accessible images, as well as contributions from prominent organizations such as the Smithsonian Conservation Biology Institute, the Wildlife Conservation Society, the North Carolina Museum of Natural Sciences, and the Zoological Society of London.

Classification Capabilities

SpeciesNet is capable of classifying images into over 2,000 distinct categories. These categories encompass specific animal species, broader taxonomic groups like “mammalian” or “Felidae,” and even non-animal objects such as “vehicle.”

This broad classification ability enhances the utility of the model for diverse research applications.

Open Source Availability and Impact

According to a blog post released on Monday, Google believes the release of SpeciesNet will empower developers, academics, and startups focused on biodiversity to enhance their monitoring efforts in natural environments.

The model is accessible on GitHub under an Apache 2.0 license, allowing for commercial use with minimal restrictions.

Alternative Open Source Tools

It is important to note that Google’s SpeciesNet isn’t the sole open-source solution for automating camera trap image analysis. Microsoft’s AI for Good Lab also provides PyTorch Wildlife, an AI framework offering pre-trained models optimized for animal detection and classification.

PyTorch Wildlife presents another valuable resource for researchers seeking automated analysis tools.