Facebook introducing the Newest Shopping AI
For those of you have ever daydreamt of being able to click photo and buy it, whether it’s your favorite celebrities’ suit or the things that they use very often, a giant social media company Facebook has proclaimed it’s moving ahead toward e-commerce game with a brand new and unique shopping experience that is powered by AI.
On May 19, 2020, in a Facebook Live, Mark Zuckerberg (Founder& CEO of Facebook) has made a statement a snap-and-shop feature that will allow users to purchase, sell and identify products seamlessly all over the Facebook platform. An auto-tagging feature – is coming on the Facebook page very soon, now autofill product description upon uploading, facilitate it to rank in search and share or exchange information in all languages. The company has the intention to make all images shoppable across its app, and in time, videos too.
Though Facebook’s new shopping AI doesn’t have users-facing names like Alex, the co-founder Sean Bell named it GrokNet to represent the scale that can analyze data very frequently. Actually, created at GrokStyle, a start-up co-founded by Bell and invested by Facebook, this universal AI model leverages computer version to recognize fine-grained product attributes from the dozens of images.
On Facebook Ait blog, the Bell has explained about how their GrokNet’s product work simultaneously analyze billion of disparate verticals like a car, clothing, and home goods. “We built, trained, and deployed a model with 83 loss functions across seven data sets to combine multiple verticals into a single embedding space. For some of the tasks that are really difficult, you need to apply a very large weight on the loss function and you need to apply many images for every batch when training. For easier data sets, you can use a small weight, and just a very small number of images per batch. We then scaled this up across all different data sets and then figured out a way of how to combine everything at once, without compromising on any one of the tasks. We use weekly supervised learning, where a model is arranged in a feedback loop to automatically create additional training data.”
As we all know that Snap-and-shop features have been using for years, by several different big companies like Amazon, Shopify, and Stitch Fix. However, Facebook’s AI team has created advanced segmentation, detection, and classification abilities that allow it to do the seemingly impossible, that recognize unseen products. That means, Facebook AI can recognize a blazer under a t-shirt, and even skirt seated under a desk. In order to get this experience, The Shopping AI modal uses a new technique known as “Instance Mask Projection” that is capable to parse clothing and segment street scenes including small details. To ensure accuracy of the forecast and account for a verity of images, Facebook’s team skilled its AI on varying people styles, ages, sound, skin, body structure, location, socioeconomics classes, and many more.
“We’re building a system that can look at a picture of your outfit, segment each piece of clothing that you’re wearing, and organize them into a digital close, allowing you to shop anything in the context of what you won,” said Tamara Berg (Facebook AI Researcher). In Future iterations, Facebook is hoping to launch an AI lifestyle assistant that helps users in different areas like AR try-outs to share with friends, wardrobe management, personalized recommendations of products to purchase, and many others. Facebook has been testing augmented reality with advertising brands such as Ray-Ban, NYX, and NARS that allow consumers to try sunglasses, lipstick, and other things virtually.
All product’s images will be uploaded in 3D with rotating views and Facebook Shops and e-commerce allow businesses to set up Facebook and Instagram storefronts for free, Also WhatsApp integration coming very soon. For those of you have been waiting to transact on the platform.
“We’re hoping to play a role in helping small businesses get through this difficult time and thrive long term,” said Mark.
According to Statista.com, With over 2.6 billion monthly active users across the world in the first quarter of 2020. The world still under the COVID-19 pandemic restriction. In such a situation, AI-powered social commerce may be a silver lining for Facebook and its users.