Machine Accurately Predicts Fashion Trends

Catwalk trends from the street be affected or vice versa? Answers provides a research team from Taipei, which has developed a mode algorithm and thus to compare trends, analyze and can predict now.

Our life is becoming more and more determined by highly complex machines and algorithms. Why so feed not a machine with data, which give us the latest fashion trends at the end and are above what we’ve put on tomorrow? Researchers of the University Taipei succeeded in now. You have developed a system for machine vision, with which inter alia to analyze is how Catwalk trends affect our wardrobe.


The question of which stood the scientist Kuan Ting Chen and his colleagues: How do affect the trends from the catwalk next season the fashion on the street? To explain that’s not possible unfortunately without mathematics.

The researchers from Taipei actually could see similarities at the catwalk trends and later Streetstyle look with its mode analysis.

For their trend algorithm, a computer was trained to recognize the posture as well as the individual regions of the body – so the arms, legs, torso, etc. – also skin, hair, colors, and textures. Two photo databases were created in the next step. The first contains 8000 catwalk looks at New York Fashion Week from the spring/summer season 2014 and 2015. In the second database are approximately 1000 Street style photos of the same period. The photo databases using the previously “trained” properties of the body were compared for the analysis. This comparison is necessary to reliably analyze whether fashion shows actually influence the style on the road.

2014: The trend towards long sleeves

The analysis showed, for example, that classic colors such as Black , white and gray as a catwalk trend coming. Also button plackets, tank tops and strips on the catwalks were widespread season 2015, in blue, cyan and red, extremely popular among designers.

Amazing that indeed similarities of the catwalk trends later reflected on the road. “We can say that a lot of people tend to imitate the style of the fashion shows,” Chen and colleagues in her essay write. But differences can be found. So a trend towards longer sleeves and bottoms factories settled in the street fashion 2014 determine – in contrast to the catwalk fashion. The researchers attributed this discrepancy to a cool summer.

Numbers game?

“Clear that fashion trends actually are a reference for the customers of fashion shows and exert considerable influence on their lives seems”, the researchers summarize their findings. We knew have always felt. But: For the first time can be expressed this belief in numbers as well.

Still, the study of the Taipei scholars is only a first basis. Would you feed the computer databases with even more photo data, these results for fashion brands, designer and trend researchers could be extremely important. Finally could be said as future sound advance, which fashion trends at the consumer is really needed, and what will sell as a result well and what is not. The purchasing behavior of customers could thus specifically controlled. Finally, no one wants to design shopkeeper and produce. Only the creative diversity could go play the flute with all statistics.