Artificial intelligence (AI) and machine learning are no longer story lines for science fiction. They are very much a part of our daily lives. Especially visual search and image recognition. Players like MadStreetDen, SnapShopr, and now Streamoid Technologies’ piQit Fashion have brought AI into our everyday fashion shopping.
Founded in 2013, Streamoid Technologies began operations in mid-2014 and is now present in India and the US. Their platform piQit Fashion uses a combination of human and machine intelligence to solve many problems commonly found in fashion retail. The team believes that blindly applying current AI techniques to a creative and constantly changing field like fashion is not enough.
Bringing in the expert voice
piQit Fashion essentially works as a fashion personalisation, analytics, and predictive recommendation platform with deep learning that is augmented with stylists’ input. “In our opinion, sophisticated AI techniques of today can solve 80 per cent of the problem, but expert guidance is required for the remaining 20 per cent,” adds 30-year-old Rajesh Kumar, CTO Streamoid Technologies.
The startup was founded by Sridhar Manthani, who was a part of initial team in S3, a computer graphics startup, which went IPO on NASDAQ. He also co-founded Thinkit, which was later sold to Intel. Rajesh, who is part of the core founding team, has also worked at Yahoo R&D and inMobi.
The visual search on the platform is able to identify the different items present in any image and tag the same basis style and category. The deep learning algorithm then searches the top brands and retailers for product images to index using the visual search algorithms. With this big data set, they then apply deep learning techniques to help generate predictive recommendations.
“We have a team of expert stylists, who are constantly rating outfits in our database and picking up where our deep learning is insufficient. This constant feedback loop ensures that our algorithms are always improving,” adds Rajesh.
Looping the customer
There is also a mobile app that users can download to organise their wardrobes. Consumers can create outfits and see how they would look wearing them. The platform in turn uses the clothes in the users’ closet to give them personalised recommendations of items they may want to add.
One can also save purchased items directly to their piQit closet from any of the team’s partner retailers with a simple SaveToCloset button that appears on the purchase confirmation page.
Essentially, piQit users are able to save clothes from all of the websites they shop at into one centralised location, and each time they do this, the recommendations they see also improve, says Rajesh. The mobile app ties directly into the retail platform to offer a multi-channel experience.
The platform has an interactive panel, linked to the retailers’ website giving the option of similar products, outfits, and styling tips. The team claims that thanks to their consumer-centric digital closet, they are able to provide analytics and user insight to retailers.
The team has already tied up with players like ABOF, Trendin (Aditya Birla Group), Pothys, Biba, and Mustard. It is interesting to note that visual search and image recognition, is fast growing not only in India but also across the globe.
A changing product
While funded players like MadStreetDen and SnapShopr are entering into tie ups with e-commerce players and giants, international players like Pinterest have already been in the space for two years and was acquired by VisualGraph. According to a report by MarketsandMarkets, this space is estimated to touch $29.98 billion by 2020, globally, with a CAGR of 19.1 per cent.
Streamoid’s first production customer signed up in mid-2015. Since then, they have added six other customers who are ‘live” with their products. Rajesh adds that every month they are adding close to two to three new customers apart from maintaining and improving their existing customer services).
“All our customers are seeing a significant increase in conversions using our product. In fact, a few of our customers thought the metrics were wrong because the results were so unimaginable. We have structured our revenue based on the value the customer derives from us and have been seeing good top-line growth since the beginning of the year,” adds Rajesh.
The team has already got two patents on visual search. They also have a patented method for automated personalised outfit recommendations. The team follows a SaaS-based revenue model that works on subscriptions, which is determined basis the value that is given. They raised a funding of $1 million by US angels last year.
“We will be launching a consumer app (piQit) to complement our B2B SaaS. This ties our technology together to help consumers across all channels – on mobile and the Web. We are going to offer our products and services to customers in other countries as well,” says Rajesh.
- machine learning
- Rajesh Kumar
- Retail and Fashion
- Image search and recognition
- Sridhar Manthani