AI for Fun: Machine Learning Makes Caricature Faces

2018-12-13

00:00 / 00:00
复读宝 RABC v8.0beta 复读机按钮使用说明
播放/暂停
停止
播放时:倒退3秒/复读时:回退AB段
播放时:快进3秒/复读时:前进AB段
拖动:改变速度/点击:恢复正常速度1.0
拖动改变复读暂停时间
点击:复读最近5秒/拖动:改变复读次数
设置A点
设置B点
取消复读并清除AB点
播放一行
停止播放
后退一行
前进一行
复读一行
复读多行
变速复读一行
变速复读多行
LRC
TXT
大字
小字
滚动
全页
1
  • A team of machine learning developers has created a system for making caricatures of people's faces.
  • 2
  • In a caricature, the artist creates a drawing of a face, with different parts of it greatly enlarged, or exaggerated, in other ways.
  • 3
  • The pictures or images are usually made to make the person look funny.
  • 4
  • Such drawings can be difficult for machines to produce.
  • 5
  • This is often because the human face is made up of complex shapes with a lot of extremely small details.
  • 6
  • Generally, good caricatures require skilled artists who can effectively exaggerate faces, while still staying true to life.
  • 7
  • The new system for making caricatures is the creation of computer scientists from the City University of Hong Kong and Microsoft Corporation.
  • 8
  • They presented their findings at a recent computer graphics conference in Tokyo.
  • 9
  • Kaidi Cao led the research. He is currently studying computer science at Stanford University in California.
  • 10
  • He helped develop the system while working as a volunteer at Microsoft.
  • 11
  • Other members of the team were Lu Yuan of Microsoft and Jing Liao of the City University of Hong Kong.
  • 12
  • Cao describes the process as a "photo-to-caricature translation."
  • 13
  • He says other machine systems for creating caricatures depend on a series of rules, which are based on how people would draw.
  • 14
  • But his team's system uses machine learning to create the caricatures from thousands of examples made by experienced artists.
  • 15
  • In a paper explaining his research, Cao said other machine-based systems appear to pay more attention to appearance and style.
  • 16
  • But his team's research uses a second machine learning tool to add "geometric exaggeration."
  • 17
  • The new method improves on existing ones "in terms of visual quality and preserving identity," Cao wrote.
  • 18
  • In addition, the system gives users more ability to change results in both shape exaggeration and appearance, he added.
  • 19
  • The team tested how recognizable its caricatures were to human test subjects in two separate studies.
  • 20
  • In the first study, people were shown caricatures and then asked to choose the correct one from a group of images with faces that looked very similar.
  • 21
  • Test results showed that in many cases, people had difficulty choosing the correct image when the caricatures were created with the older methods.
  • 22
  • The researchers said this was because such systems often created caricatures that were far too exaggerated or unclear.
  • 23
  • Not surprisingly, the most easily recognized caricatures in the experiment were made by humans.
  • 24
  • But the researchers said their system performed better in tests with human subjects than the older methods.
  • 25
  • The second study tested how close the system's caricatures were to hand-made caricature styles.
  • 26
  • People were first shown eight caricatures drawn by artists.
  • 27
  • They were then presented with one hand-drawn caricature and five machine-created examples.
  • 28
  • Test subjects were then asked a series of questions about which examples were the most similar.
  • 29
  • Results showed that people taking part in this experiment chose the team's machine-created caricature about 23 percent of the time.
  • 30
  • This means the system "can fool users into thinking it is the real hand-drawn caricature," the researchers wrote.
  • 31
  • The team noted, however, that "an ideal fooling rate" for such a system would be about 50 percent.
  • 32
  • To date, the team has mostly worked on developing caricatures of people's faces.
  • 33
  • But the researchers say they plan to expand on this in the future to include full bodies and other objects.
  • 34
  • They also plan to explore turning the current model around to permit caricature-to-photo translation.
  • 35
  • The team says it believes this might be useful for facial recognition of caricatures.
  • 36
  • I'm Bryan Lynn.
  • 1
  • A team of machine learning developers has created a system for making caricatures of people's faces.
  • 2
  • In a caricature, the artist creates a drawing of a face, with different parts of it greatly enlarged, or exaggerated, in other ways. The pictures or images are usually made to make the person look funny.
  • 3
  • Such drawings can be difficult for machines to produce. This is often because the human face is made up of complex shapes with a lot of extremely small details. Generally, good caricatures require skilled artists who can effectively exaggerate faces, while still staying true to life.
  • 4
  • The new system for making caricatures is the creation of computer scientists from the City University of Hong Kong and Microsoft Corporation. They presented their findings at a recent computer graphics conference in Tokyo.
  • 5
  • Kaidi Cao led the research. He is currently studying computer science at Stanford University in California. He helped develop the system while working as a volunteer at Microsoft. Other members of the team were Lu Yuan of Microsoft and Jing Liao of the City University of Hong Kong.
  • 6
  • Cao describes the process as a "photo-to-caricature translation." He says other machine systems for creating caricatures depend on a series of rules, which are based on how people would draw. But his team's system uses machine learning to create the caricatures from thousands of examples made by experienced artists.
  • 7
  • In a paper explaining his research, Cao said other machine-based systems appear to pay more attention to appearance and style. But his team's research uses a second machine learning tool to add "geometric exaggeration."
  • 8
  • The new method improves on existing ones "in terms of visual quality and preserving identity," Cao wrote. In addition, the system gives users more ability to change results in both shape exaggeration and appearance, he added.
  • 9
  • The team tested how recognizable its caricatures were to human test subjects in two separate studies.
  • 10
  • In the first study, people were shown caricatures and then asked to choose the correct one from a group of images with faces that looked very similar.
  • 11
  • Test results showed that in many cases, people had difficulty choosing the correct image when the caricatures were created with the older methods. The researchers said this was because such systems often created caricatures that were far too exaggerated or unclear.
  • 12
  • Not surprisingly, the most easily recognized caricatures in the experiment were made by humans. But the researchers said their system performed better in tests with human subjects than the older methods.
  • 13
  • The second study tested how close the system's caricatures were to hand-made caricature styles. People were first shown eight caricatures drawn by artists. They were then presented with one hand-drawn caricature and five machine-created examples. Test subjects were then asked a series of questions about which examples were the most similar.
  • 14
  • Results showed that people taking part in this experiment chose the team's machine-created caricature about 23 percent of the time. This means the system "can fool users into thinking it is the real hand-drawn caricature," the researchers wrote. The team noted, however, that "an ideal fooling rate" for such a system would be about 50 percent.
  • 15
  • To date, the team has mostly worked on developing caricatures of people's faces. But the researchers say they plan to expand on this in the future to include full bodies and other objects.
  • 16
  • They also plan to explore turning the current model around to permit caricature-to-photo translation. The team says it believes this might be useful for facial recognition of caricatures.
  • 17
  • I'm Bryan Lynn.
  • 18
  • Bryan Lynn wrote this story for VOA Learning English. George Grow was the editor.
  • 19
  • What are your thoughts on these machine-created caricatures? Write to us in the Comments section, and visit our Facebook page.
  • 20
  • _______________________________________________________________
  • 21
  • ________________________________________________________________
  • 22
  • Words in This Story
  • 23
  • caricature - n. a funny drawing of someone that makes part of their appearance or character more noticeable than it really is
  • 24
  • exaggerate - v. make something seem larger, better or worse than it really is
  • 25
  • translation - n. process of changing from one form to another, such as with languages
  • 26
  • style - n. way of doing something that is typical of a particular person, group, place, or period
  • 27
  • geometric - adj. of, relating to, or according to the methods or principles of geometry
  • 28
  • preserve - v. to keep something the same or prevent it from being damaged or destroyed