2016 has been the most important year yet for Deep Learning/AI. These fields had predominately been more academic and research oriented by virtue. We at Talespin, In the past year have worked extensively on deep learning in Fashion.

Implementations of Deep Learning in the past year have many success stories. But is the idea to use neural networks to model data really new?

Many argue that this has been there for years and everything now is the phase called hype.

In a sense they are right. The logical implication has been around for a while. But what is now available are affordable hardware, infrastructure, open source libraries and vast amounts of data. The buzz for the past few years has been Big Data. Now with the powers of deep learning enterprises have started training systems to make sense out of it.

Here is an extract from a brilliant article by Mikhail Naumov for forbes.

While applicable to an endless number of use-cases, this method follows some general principles in order to be practical and achievable in the near term:

  1. A company must have lots of historical data to train the deep learning algorithm
  2. A company should have a recurring need for predicting things that either:
  • Cut costs: for example reducing average handling time in a customer service conversation; or reducing the need for in-person insurance assessments
  • Create value: like up-selling the right product to the right customer at the right time; or helping marketers create engaging content which will lead to more sales

The requirements above point to a number of business use-cases, which are going to see major transformation over the next 12 months.

Business Use Cases for AI-powered Transformation:

  • Customer Service: Currently the largest market, and exceptionally well-positioned for disruption, due to availability of vast historical customer service data. Our company — DigitalGenius — operates here bringing meaningful results to companies like KLM Royal Dutch Airlines and partnering with leading platforms like Salesforce Service Cloud & Zendesk.
  • Sales: Another obvious use case. Just think of all the emails you get inside your inbox from people trying to sell you something. Very soon, those emails will be hyper-personalized, and will only land in your inbox during the 20-minute time slot when you’re statistically most likely to open them and respond positively. Salesforce Einstein, for example, makes it easy for sales professionals to focus their time on the most important leads, through predictive lead scoring.
  • Marketing: Marketers have one major pain point today — too many data elements to segment, organize and learn from. They are suffering from data overload thanks to an endless menu of analytics tools. In 2017, deep-learning algorithms will bring order to their marketing data and provide real-time recommendations for audience targeting, campaign timing and content marketing. For good examples check out Radius and Persado.
  • Operations: Companies like x.ai are already achieving near-perfect automation of meeting scheduling. And in 2017 will likely become household names inside medium and large enterprises. Similarly, recruitment chatbots like Mya will screen candidates and handle all communication with prospective talent, Saving companies time & valuable resources in the talent acquisition process. Tools like Clarke.ai will dial into our conference calls and send a summarized outline with action-points and to-do lists to all the participants afterwards.
  • Government Affairs: Notably, more sensitive areas like government affairs will finally become transparent and preemptively actionable. For great examples look at the way FiscalNote analyzes government data to predict outcomes of law-making processes around the country.

Coming to fashion. Where all do we feel Deep learning can have in impact?

  1. Images Processing — Deep Learning offers exciting opportunities in “understanding” and organizing huge amounts of images.

    This learning can be very helpful in taking a learned call towards various business decisions

    • Sourcing
    • Designing
    • Trend analysis
    • Customer sentiment analysis
  2. Attribute tagging — Prodcut database can bee used to train models to classify images. Using deep learning libraries you can train models to capture the general attributes of clothes (like shirts, pants, shoes or even collar, sleeves etc).
  3. Recommendation Engine — Deep learning can be used to suggest looks and styles using a specifc product catalogue and a consumer’s interest.
  4. Fashion Discovery — Just visit any famous fashion facebook group or Instagram handle. Majority of the queries are “Where can I buy this?” or “This celebrity was wearning this dress on a specfic event. Where can I buy a similar dress?”

Brands can structure their catalogue as a visual search engine. Let a user upload a picture of dress and find similar products from the brand catalogue.

At Talespin, we are trying to solve many of the above mentioned use cases. All of this as a plug and play solution without any huge infrastructure cost.

We will be posting more use cases and case studies of deep learning in fashion soon.

Similar Posts