Fashion AI in action, visual and bot

Part of the game of fashion business is about image or photo, usually tons of them. That is why the surge of artificial intelligence has been developing automatic&real time image recognition for any photo posted on the internet. For some startups, visual and bot solution are becoming the new alchemy. 


  • Image recognition to detect trend

This year a French startup Heuritech beat out other 500 candidates to win LVMH’s First Innovation Award at Viva Technology Event in Paris. It claims to leverage machine learning to identify hundreds of details inside fashion images, then forecasts buzz trends related to shapes, patterns, colors, shades, style etc. It is said the startup is already working on some pilot projects with LVMH. We actually noticed several similar startups which are trying to predict trends for designing collections, optimize merchandise, and make qualitative or quantitative customer segmentations, all based on their AI powered image recognition capabilities. 



  • Visual search for personalization&recommendation

Backed by Sequoia,, a sub brand of Indian tech startup Mad Street Den, enables retailers to personalize the e-commerce according to individual shoppers’ preferences. The technology learns from shoppers’ browsing and shopping histories and displays items that match their preferences and provides personalized shopping recommendations. It uses AI and tagging to showcase similar products, personalized homepages for each shopper, tailored product recommendations and much more. It is now serving Indian e-commerce sites such as Craftsvilla, Voonik, YepMe and told its service generates an increase of 144% in product viewing and a 110% increase in average order value.





  • Fashion bot

The chatbot is every where. For example, British e-retailer Shop Direct is working with IBM Watson to develop AI chatbot that can perform sentiment analysis to detect a shopper’s mood, based on word usage and message tone. This approach is about helping customer service support, freeing its workforce from repetitive tasks and focusing more on the interactions that really help customers.


On the other hand, more and more startups are looking into making fashion bot that functions as a virtual assistant for visual search and styling inquiries. is one startup we picked here as an example. It provides a good demo to explain its technology (see below video). Basically, it uses a bot solution (can be easily integrated with major Messenger platforms) to recognize the item of clothing in the photo you upload, show you the exact match or visually similar match, and direct you to the purchase site under your own budget request. The Facebook Messenger bot launched in March this year. If you send the bot a picture,  it will give you three options: More Like This, Buy, and Who Wore It.



  • Takeaway

For many Chinese, visual search is not a cool new tool. Alibaba’s e-commerce platforms Taobao and Tmall have implemented visual search function a long time ago. It works pretty well, nevertheless, not up to perfection. For instance, we uploaded the image of Selena Gomez outfit on Tmall:




Below is the result: top sweater seemed to find the closest match, as for pants and handbag, not that ideal.

We have observed lots of startups, whether from west or east, are actively diving into this space lately, believing improved AI algorithm should be the new answer to enhance accuracy and efficiency for visual search/recommendation in the fashion. The competition is actually getting very intense because internet giants like Google, Facebook, Microsoft, Adobe, IBM, Alibaba, Tencent also have AI services to power computer vision. In some extent, it demonstrates the fashion industry will and should gradually adopt above technologies to generate precise customer insights while driving engagement and sales.


Author: Cecilia Wu