Behavioral analysis can reveal a lot about an animal’s health status or motivations, but it can be challenging to accurately interpret the movements of various species. A new technology developed at EPFL called SuperAnimal uses a single deep learning model to detect animal motion across a wide range of species and environments. This groundbreaking “foundational model” can be applied in animal conservation, biomedicine, and neuroscience research to provide valuable insights into animal behavior.
Even though animals can’t verbally communicate their physical or emotional states, their movements can provide important clues. Pain or joy can manifest in changes in gait or facial expressions, which can be easily observed by a trained eye. By automating animal behavior analysis with AI models like SuperAnimal, observer bias can be eliminated, and humans can more efficiently access important information about the well-being or motivations of animals, including cows, dogs, cats, and mice.
A notable advancement in posture analysis for behavioral phenotyping, SuperAnimal is an open-source tool developed by Mackenzie Mathis’ laboratory at EPFL. This innovative tool can automatically recognize the location of “keypoints” such as joints in over 45 different animal species without the need for human annotations. By standardizing the process and making labeling more effective, SuperAnimal revolutionizes motion analysis by providing a more efficient and precise method for tracking and analyzing animal behavior.
The SuperAnimal method builds upon a pose estimation technique previously developed by Mackenzie Mathis’ laboratory, known as DeepLabCut™. By compiling a large set of annotations across databases, the model learns a harmonized language and can be easily deployed or fine-tuned by users for further customization. This approach simplifies the process of training large-scale datasets and makes motion analysis more accessible to a wide range of users, including veterinarians, biomedical researchers, and those studying laboratory mice.
The potential applications of SuperAnimal are vast, extending beyond traditional fields such as veterinary medicine and biomedicine into neuroscience and even athletic performance analysis. Future developments aim to expand the model’s capabilities to include other species like birds, fish, and insects, as well as integrating natural language interfaces for more accessible and advanced behavioral analysis tools. SuperAnimal is now available for researchers worldwide through open-source distribution, empowering scientists to utilize this cutting-edge technology for a variety of research purposes.
In conclusion, SuperAnimal represents a significant advancement in the field of animal behavior analysis, providing researchers with a powerful tool to track and analyze animal motion across multiple species and environments. By automating the process with AI technology, SuperAnimal eliminates observer bias and offers a more efficient and standardized approach to studying animal behavior. With its wide range of potential applications in animal conservation, biomedicine, neuroscience, and beyond, SuperAnimal is poised to revolutionize the way researchers study and understand the movements and motivations of living beings.
