Face recognition of individual chimpanzees from Bossou, Guinea. Credit: Kyoto University, Primate Research Institute
While most of the current stories surrounding facial recognition technology sound like something from the realms of dystopian science fiction, as valid concerns about how certain programs could be developed to infringe upon civil liberties, researchers at Oxford University have discovered a more worthwhile use for the technology.
Scientists at Oxford have developed new cutting-edge software driven by artificial intelligence that can detect and track the faces of individual chimpanzees in the wild. This new deep-learning algorithm could assist researchers and wildlife conservationists as they come to understand animal behavior better. Furthermore, the technology will save valuable time and resources spent on tracking the animals and analyzing video footage.
The Oxford team trained the AI by taking around 50 hours of footage recorded over 14 years and feeding it directly into a deep neural network. This amounted to around 10 million facial images of 23 chimpanzees, with age ranges estimated to range from newborn to 57-years old.
Access to this large video archive has allowed us to use cutting edge deep neural networks to train models at a scale that was previously not possible.
Arsha Nagrani, DPhil Student at the Department of Engineering Science, University of Oxford
The resulting algorithmic model had the ability to identify individuals with up to 93% accuracy and correctly categorize the sex of the primate close to 96% of the time. Comparatively, the algorithm also outperformed expert human labelers given nearly an hour to complete the task – the facial recognition system, in contrast, took just a fraction of a second.
Although facial recognition tools have previously been used to track animals – ChimpFace is a tool currently used to fight illegal trafficking of chimps. The researchers of this latest breakthrough say their system also outperforms its predecessors by significantly reducing the processing required on raw footage, as Nagrani stated, “Additionally, our method differs from previous primate face recognition software in that it can be applied to raw video footage with limited manual intervention or pre-processing, saving hours of time and resources.”
The software also displayed its high-performance in trials working well on low-light and poor-quality images as well as instances when the chimps weren’t directly facing the camera.
The potential for the technology does go beyond the scope of behavioral studies and fighting the illegal movement and trade of chimpanzees as the software could be applied to a broad range of species. Thus, by utilizing the program for the conservation studies of other animals it could help drive further applications of artificial intelligence to solve many of the problems across the wildlife sciences, such as endangerment and monitoring the biodiversity of a particular environment.
The software is open source and available to the wider research community as stated in the paper, published in ScienceAdvances. “We hope that this will help researchers across other parts of the world apply the same cutting-edge techniques to their unique animal data sets. As a computer vision researcher, it is extremely satisfying to see these methods applied to solve real, challenging biodiversity problems,” says Nagrani.
Therefore, by harnessing the power of artificial intelligence and machine learning the team hope to alleviate the pressure put on working conservationists by providing with the tools they need for accurate long-term behavioral studies and the safeguarding of various wildlife species. The interactions and movements change over multiple years and generations of animals, says Daniel Schofield, researcher and DPhil student at Oxford University’s Primate Models Lab, School of Anthropology, “You can start to build up a social network,” he says.
With an increasing biodiversity crisis and many of the world’s ecosystems under threat, the ability to closely monitor different species and populations using automated systems will be crucial for conservation efforts, as well as animal behavior research. Interdisciplinary collaborations like this have huge potential to make an impact, by finding novel solutions for old problems, and asking biological questions which were previously not feasible on a large scale.
Daniel Schofield, Researcher and DPhil Student at Oxford University’s Primate Models Lab, School of Anthropology