With the arrival of the trend of full-screen mobile phones, many people have experienced face recognition technology through mobile phones. In fact, facial recognition not only makes our life more convenient, it can also be used on animals, and even save some endangered species, "beast face recognition" to understand?
Recently, researchers at Michigan State University in the United States have developed PrimNet , a facial recognition system for primates that can track these animals with Android applications to help forest conservationists better protect some endangered primates.
According to the project the researchers Anil Jain introduction , the main use of facial recognition system PrimNet convolution neural networks and artificial intelligence technology, a Cornell University showed that, PrimNet recognition accuracy in the test than the current majority has to face recognition high.
Currently, this system is mainly used to identify endangered primates such as gorillas, gold-tailed monkeys, and lemurs. Although these animals and humans are primates, facial recognition of animals is very different from that of humans.
Like many neural network systems, researchers also use a large number of photos as a sample to train PrimNet, but the difficulty is much greater than the face. Researchers have used face recognition technology in these primates and there has been an "overfitting" problem, which means that the details are not obvious.
This is because the characteristics of the animal's hair and eye color are quite different, so the research team had to manually mark the animal's eyes and mouth to assist in system identification.
In practical applications, the nurse can upload the captured animal photo to an Android application named Prim ID. The system displays the relevant data that matches the animal or gives the five closest options.
At present, the recognition accuracy of the PrimNet system exceeds 90%, but this is not ideal for facial recognition. Therefore, there is a trade-off option of “giving 5 similar options.” The researchers stated that they are working hard to improve the recognition accuracy.
Although PrimNet's performance is far from satisfactory, once the recognition rate is further improved, it will be a cheaper and more efficient animal tracking method. The price of current animal tracking devices is mostly between US$400 and US$4,000. It needs to be captured at the same time. Animals implanted in the body to track the equipment, but also easy to cause harm to animals.
(Investigators implanted a solar-powered microsatellite tracker in a giraffe horn, from: National Geographic China Website)
The researchers hope that the facial recognition system can be applied to more primates. The PrimNet system has been open-sourced and more people have been invited to join them.
In fact, "face recognition" is not new in the field of facial recognition. Last year, researchers at the University of Cambridge developed a sheep facial expression recognition system to determine the degree of pain in sheep. When the sheep is in a state of illness and injury, Ranchers can better understand the status of sheep.
In China, in April this year, Baidu announced the official launch of the System of Animal Facial Identification SAFI (System of Animal Facial Identification), but this system can only be used for dog face recognition.
SAFI has been installed in pet food bowls, access control, courier, and payment systems. It can be used to feed the dog's face, enter and exit the entrance, collect express delivery, and pay for goods.
In March this year Professor Chen Yaosheng of Sun Yat-Sen University also officially released the "Pig Face Recognition " technology. The identification rate of sows is 98%, and the identification rate of hogs is 85%. It can be identified by "skin brushing" according to each pig in the breeding process. The state has allocated different feeds. There have been 168,821 large pigs in 1,692 pig farms across the country involved in pig face recognition tests.
Maybe one day, the comet of your family can brush your face to unlock your phone.
0 comments:
Post a Comment