In the 2006 film the Devil Wears Prada Queen (also known as The Devil Wears Prada), fashion magazine editor in Chief of Milanda·puruisite just at a glance, can be analyzed, who designed clothes for others, and know where clothing design inspiration can be traced back to one year. Puruisite, a role inspired by Anna Wintour, her longtime fashion editor of Vogue magazine, which itself is synonymous with fashion.
Humans can pass a simple glance, identification and confirmation of fashion designers and fashion style, so why can't a machine do that? Today, Taiwan National University, KuanTing Chen and several partners shows that the machine could do this, machine vision can distinguish between one quarter to the next fashion changes. In particular, they use machine vision technology can be shown after the fashion show on the impact of street clothing.
Chen and its partners a began through training machine Visual algorithm, recognition pictures in the models of body posture, then will body divided into nine a different of regional--arm, and Xia arm, and thigh, and Shank (distinguish around limb) and torso, analysis these regional of clothing color, and texture and skin, information to created a list, its role similar Yu for whole body created a Visual features aspects of vector list.
So, fashion style became more dimension of vector, this relatively simple mathematical process.
Later, they will collect two sets of databases. The first group contains 8000 in 2014 and 2015, Zhou Chun/New York fashion summer fashion model pictures. Group II are contained in 2014 and 2015, taken during the spring/summer of about 1,000 street fashion clothing pictures.
Fashion week in New York is a very important activity, which includes about 300 events, and can attract up to 100,000 people to attend. Therefore, an interesting question is held this week emerging trends in activity, in the next season will affect street fashion. To find the answer, Chen and his team of machine vision algorithms to identify these trends.
Results are very interesting. Algorithm discovered a small amount of classic clothing styles are common in every year. Team, whether it be in 2014 or 2015, the upper part of the costumes in the fashion images in use a lot of white, grey and black.
However, most of the key points of interest is to understand how changes in fashion, the algorithm can help. Survey found that in 2015, the more popular styles include placket shirt and Camisole, and striped clothing materials, more popular colors are blue, green and red.
Funny thing is, this change in fashion in the popularity also reflects the data presented on the street fashion. Team said data can tell you, a lot of people are imitating styles shown in the fashion show.
However, the fashion week shows trends and street wear, there are some differences between the data sets. For example, the team found that the 2014 in street fashion clothing, knee-length trousers and long-sleeved clothing's popularity rose. Team speculated that this should be compared to other years to cold in the summer of that year, so people will wear slightly warmer clothing covering the body. Huawei is not only a fake master is a skilled
Meanwhile, it also shows the influence of culture, such as how quickly the fashion throughout the community. Message from fashion shows fashion trends are indeed provide a reference to clothing, and has a significant impact on people's daily lives.
Many fashion followers are very familiar with it. The difference is now quantifying its influence to methods in computer vision and fashion stores, designers and consumers provide valuable feedback. This may be a way of using machine vision technology profits.