Thu. Jan 23rd, 2020

New use of facial recognition: apply to chimpanzees to better understand them | Innovation

Over the past few years, facial recognition has developed a bad reputation because it poses a risk to user privacy. It is already used in China, as a curious fact, to embarrass pedestrians who cross a zebra crossing in red (in doing so, their face appears on a luminous sign visible to everyone). The Big Brother who watches also collects data to make more weighty decisions, how to grant or not mortgages or a job. But outside the human realm, a team of researchers from the University of Oxford has discovered a new, more laudable purpose for this technology: to help them monitor the behaviors and interactions of chimpanzees to improve the understanding that scientists have of these animals and of Your social interactions. It could also be used to improve existing efforts to track endangered species.

The study, published in the scientific journal Science Advances Earlier this month, he explains that the deep learning algorithm is trained to recognize the sex and identity of wild chimpanzees. To achieve this, they used approximately 50 hours of images taken over 14 years. "The scientists extracted 10 million facial images of 23 chimpanzees and introduced them into a deep neural network. The resulting model was able to identify individuals with up to 93% accuracy and classify their sex correctly up to 96% of the time," explain in MIT Technology Review.

When comparing the machine's performance with that of humans, the researchers found that the model was twice as efficient as the labeling experts who needed almost an hour to complete the task, and four times better than the rookies who needed almost two hours. The model, in contrast, only needed a fraction of a second. Of course, the few times he failed, he confused a chimpanzee's butt with a face.

This new use of face recognition, made possible by artificial intelligence, considerably facilitates the task of scientists. One of the techniques that researchers use to track animals and study the behaviors of populations over time is the use of video images. After collecting gigas and gigas of moving images, it is difficult to classify and identify and recognize each of the animals that appear in them. Classifying the large amounts of data they accumulate is tedious and time consuming. In addition, labeling them manually can lead to inaccurate results. Artificial intelligence is emerging as a promising method to facilitate these tasks and accelerate research in animal behavior.

This is not the first time facial recognition is used to track animals. ChimpFace is a similar tool that is mainly used to combat illegal chimpanzee trafficking. Other studies have tried to track lemurs, baboons and other primate species in danger of extinction. However, the algorithm of this last work, according to the researchers, improves on the predecessors because it minimizes the amount of processing required in the raw images. The previous algorithms had problems to handle lighting variations and poor image quality, the new algorithm works better in those conditions because it was trained in a more diverse set of data.

. (tagsToTranslate) new (t) use (t) facial recognition (t) (t) apply (t) chimpanzee (t) understand (t) better (t) human (t) technology (t) assume (t) threat ( t) privacy (t) be (t) useful (t) help (t) scientist (t) behavior (t) wild

Source link