Machine vision takes you to the fashion world

In the "Queen of Prada" movie, the editor-in-chief of fashion magazine Miranda Priest, you can see who designed the clothes others wear, and know which year the clothing design inspiration can be traced back to .

1528170782882647.png

Today, this can also be done completely by using machine vision. Machine vision can distinguish fashion changes from one quarter to the next, and machine vision technology can also show the impact of fashion shows on streetwear. We are here today to discuss with you the research of machine vision in the fashion industry.

First, train the machine vision algorithm to recognize the body pose of the model in the picture, and then divide the body into nine different regions-upper arm, lower arm, thigh, calf (distinguish between left and right limbs) and torso, and analyze the clothing color of these areas , Texture, and skin to create a list, similar to creating a vector list of visual features for the entire body. Therefore, the more fashionable style becomes a relatively simple mathematical process of comparing vectors of multiple dimensions.

Thereafter, at least two sets of databases were collected. For example, the first group contains 8,000 pictures of spring / summer fashion models for 2014 and 2015 New York Fashion Week. The second group contains approximately 1,000 pictures of street fashion clothing taken during the spring and summer of 2014 and 2015, as shown below.

1528170806833486.jpg

Through the algorithm, we can find that a small amount of classic clothing styles are very common every year. No matter in 2014, 2015, or even 2017, the color of the upper clothes used white, gray and black in the fashion image.

The survey found that in 2015, the more popular styles included half-cardigan pullovers and camisole, as well as clothing materials with stripes. The more popular colors were blue, cyan and red. In this regard, machine vision algorithms can also analyze changes in fashion, and the data presented in this fashion change in popularity also reflects street fashion. The data can tell you that many people imitate the clothing styles shown in fashion shows.

However, there are some differences between the trends revealed during Fashion Week and the streetwear dataset. For example, over-the-knee trousers and long-sleeved clothes have increased in street fashion in 2014. It should be that the summer of that year is colder than other years, so people will wear slightly warmer clothes to cover their bodies.

Obviously, the fashion trends of fashion shows do provide a reference for people and have a major impact on people's daily lives. Many fashion followers are familiar with this, except that its impact can now be quantified in computer vision and provide valuable feedback to fashion stores, designers and consumers, which may also be the use of machine vision A way to monetize.