Competitive Analysis — Image Recognition Technology

Image Recognition Technology

Google Style Match

  • On May 15, 2018, Google launched a new feature, Style Match, which capitalizes on the new trend of helping shoppers find exactly what they’re looking for with a snap of a picture.

  • Style Match enables Android users to use both Google’s search engine and the new Google Lens to identify and buy products. Style Match’s algorithm even understands attributes like texture, shape, and color.

  • Users simply point their phone cameras at objects and Google will search through millions of products to identify what they are and provide a match. If Style Match cannot find the exact product, they will suggest similar products.

  • Once a user finds the product they are looking for, they can purchase it directly through Google Shopping or be directed to an external site.

  • Like other image recognition apps, Style Match aims to help users find products they can’t put into words in Google Search.

Using Google Style Match to find an article of clothing in a store.

Slyce

A Universal Scanner Solution

  • Combines 3D real-world product recognition with barcode and catalog scanning technologies

  • Finds visually similar pieces result set based on the visual attributes and metadata extracted from the image

  • Used by Tommy Hilfiger, Neiman Marcus, Urban Outfitters

    • Data showed that Slyce integration encouraged more app usage, more mobile transactions, increased conversions in-store and online, enhanced customer engagement, and removed search hurdles

Slyce Technology An overview of the ideal use case.

Tommy Hilfiger: Runway Recognition

  • Enabled users to purchase the looks as soon as they saw them on the runway or in a pop-up shop (Tommyland fashion show in Feb 2017)

  • Direct pathway to Tommy Hilfiger’s eCommerce platform when a customer selected a saved look for seamless transition to point of purchase

  • Use case: A girl is watching the recorded fashion show or the fashion show in person. She sees a model wearing a really nice ensemble and wants to purchase it without having search hurdles. She opens up the Tommyland app with Slyce integration and snaps a picture of the model. The snap uploads and Slyce’s algorithm works to recognize the items in the image. Once its complete, it creates a look in her list. Shecan click on the looks and see the different products worn by the model. She can easily click on the individual products and be directed to the product page where she can get more info and purchase.

  • She can also takea picture and choose it from her library to upload onto the app

Tommy Hilfiger Snap:Shop App. Built with Slyce technology.

Neiman Marcus

  • “Snap. Find. Shop,” which was developed with Slyce, allows customers to search Neiman Marcus inventory based on every day inspiration.

  • App users can simply snap an image of an item and will be instantly provided with all similar items that are currently available at NeimanMarcus.com. Customers will not have to crop, search or select a specific category. If the customer is interested in the item, he or she will then be able to purchase the item at the exact moment.

Neiman Marcus App built with Slyce Technology

Craves (a Slyce Product)

  • See a shirt you love on Instagram, Pinterest or Tumblr? Upload a photo or screenshot of it and Craves will show you where to buy visually similar items

  • Social integration: Users can follow celebrities, brands, and other users

  • Alert system: save an item for later and Craves notifies you if the price drops

Craves An app built with Slyce technology.

Snap Tech

Snap Similar

  • Integrated with LOOK Fashion e-commerce site

    • “SNAP SIMILAR” button that lets readers and customers shop by similar size and shape

    • Users can use dots to trace the outline of the product they want to search for similar results

  • Their visual search algorithm returns results based on color or shape across your whole inventory

  • Reduces clicks through pages of unsuitable products

  • Gives users ability to find items similar to those they already clicked on

  • No meta tagging needed

  • Can be added to emails and newsletters

Snap Similar Technology

Donde

Icon-based Search

  • Ditches text search and opts instead for icon-based search, letting users simply tap on an icon for a specific color, pattern, length, etc. to find all items that match that description

  • Users don’t have to actually know exactly what they want and try to put it into words, but can instead choose from a number of options

    • Very specific features: neckline, sleeve, length, length of pant leg

    • Users can filter by brand and set a price range as well

  • Users can follow various brands and friends so that they can get style tips from folks that they trust

  • There are Sale Alerts for products users save

Donde's Icon Based Search

Amazon Echo Look

 

Amazon Echo Look

  • Purchase by invitation-only

  • Includes a lookbook feature to help you discover new brands and styles

  • Be advertised to based on what you’re already wearing

  • Takes full length photos (5 MP) and videos to capture outfit from every angle

    • Built-in LED lighting and depth-sensing camera let you blur the background to make your outfits pop, giving you clean, shareable photos

  • Doubles as a speaker and runs Alexa

  • Customers themselves can also add in their vote for different outfits in the app

    • Helps Amazon learn about you

  • Style Check feature: enables user to compare two different outfits and find out which one is “better”

    • Combines machine learning algorithsm with advice from fashion specialists

    • The existing, outdated Outfit Compare feature: judgment relies entirely on old-fashioned human input

    • Major criticism: taste is subjective, perpetuating a dated binary of a “right” and “wrong” way to dress

Amazon’s Recent Developments

  • August 2017: Amazon released their own clothing line: Find.

  • May 2017: Amazon was granted US government approval on a manufacturing patent for a fully machine-operated apparel factory with the capability to produce clothing “on demand”

  • Prime Wardrobe: lets customers try on clothes before buying them

Generative Adversarial Network (GAN)

  • Amazon researchers based in Israel developed machine learning that, by analyzing just a few labels attached to images, can deduce whether a particular look can be considered stylish

  • Uses a cutting-edge tool called generative adversarial network (GAN)

    • Developed by a researcher on the Google Brain team

    • Consists of two deep neural networks operating in tandem to learn efficiently from raw data

    • The GAN internalizes the properties of a particular style simply by looking at lots of examples, and it can then apply that style to an existing item of clothing

Epytom

Facebook Stylist-Bot

  • Epytom has analyzed outfits from style icons of the past, like Audrey Hepburn, as well as today’s top tastemakers

  • ~5 million messages/month

  • A portion of the natural language processing (NLP) that gives the bot the ability to answer questions comes from API.ai, the service for bot or intelligent assistant creation acquired by Google last fall

    • Vast majority of its NLP, however, comes from deep-learning algorithms developed by Eyptom Stylist and based on millions of questions users have asked the bot since it was created

  • Creates personalized looks tailored to your style and the weather

    • For clothes you already own (pnly requires 40 pieces)

Epytom.com's Facebook Messanger Style Bot