题型:阅读理解 题类:常考题 难易度:普通
浙江省山河联盟学校2020-2021学年高二下学期英语4月月考试卷
As missing children age, their appearances are quite different from their last known photographs, which makes finding them tough. Now AI could make it quicker to match any found children to those old photos.
Police can use software to age photos of people who have been missing for a while, but these tend to work best in adults, as age-related changes to adult faces are more predictable and so easier to simulate(模拟).
With children, these aged images don't tend to be like the older child, and matching photos of found children to old images in a database of missing children is difficult.
Now Debayan Deb, a professor at Michigan State University, and his colleagues have created an program to do this for them. They trained a face-recognition program on data sets including the Children Longitudinal (纵向的) Face data set, which contains images of nearly 1000 children aged between 2 and 18 years old. Each was photographed at least four times over a period of six years.
The AI learned to match recent photographs of children with images taken 2. 5 years earlier 80 per cent of the time.
With one year between the two photographs, the method was 90 per cent accurate at identifying faces. This dropped to 73 per cent after three years. The approach beats comparing children with photos taken when they were aged 0 to 4 that have been aged by software, which hovers at 50 per cent recognition after six months. The new AI might help to improve accuracy of this kind of software as well.
"This is a very good paper that could potentially be used to identify missing children," says Amarjot Singh at the University of Cambridge.
Deb's next goal is to extend the age gap. His team also hopes to develop an app that could be used to fight child trafficking (拐卖).
试题篮