ViSenze's mission is to make sense of the visual web
By Gabey Goh November 15, 2013
- Currently in talks with interested investors for its Series A fundraising round
- United States and China key markets in expansion plans
A STARTUP srpung from the NExT Research Centre, a collaboration between the National University of Singapore and Tsinghua University of China, is seeking to help both users and businesses make sense of an increasingly visual online world.
ViSenze, founded in 2012, has a mission to unlock the intelligence captured in any rich visual content using sophisticated visual recognition technology and visual driven analytics.
In an interview with Digital News Asia (DNA), cofounder and chief executive officer Oliver Tan (pic) shared that the work being done on the company’s proprietary technology didn’t really accelerate into real world applications until the team found real-use cases with an initial partner.
“It was pure fate that one of my cofounders, Roger Yuen, ran an online social fashion experiential site called Clozette and he immediately saw the compelling use of visual recognition and search to drive fashion e-commerce. It also happened to be the perfect platform for us to experiment and fine-tune our technology,” he said.
So a mutually beneficial relationship was established between the team and Clozette, which proceeded quietly for more than a year until they were ready to publicly announce the solutions.
ViSenze’s official market debut was via the announcement of a partnership with Rakuten, with the company’s visual search and recognition technology powering the novel Fashion Finder application on two sites focused on Taiwan within the e-commerce player’s network.
This application allows users to search and buy similar fashion items that they see on the sites, from user-generated photos and user-uploaded images, which could be taken from fashion websites or e-magazines.
Tan claimed that initial results since the partnership was announced in July have been "very positive" but preliminary, and declined to share more details, citing a confidentiality agreement with Rakuten.
He also claimed there was a “floodgate of interest” from companies seeking to utilise the solution for their own online businesses since ViSenze's debut.
“We are currently in discussions to expand our deployment to other Rakuten sites globally. ViSenze also has several confirmed projects with leading telcos, advertising networks and media owners in the region,” he said.
It is no secret that the online world is becoming more visual, with over 600 million images posted to the web everyday. As noted by Tan, the challenge for users is how to find those images they are looking for quickly, without always having to guess the keywords to describe them.
“The problem they face is that web search engines today are all designed to read text, and not pixels in images. So users still end up guessing keywords to find the images. For the visual world, where a picture says a thousand words, our solution is to help users ‘start finding with visual search and stop guessing with keywords.' This is an immensely more efficient way,” he added.
Navigating the visual web
Tan noted that visual search and recognition technology is not new, and there are other players already in the market such as Google Image search and specific visual technology like TinEye.
“But most are just image-matching without real in-image object recognition and visual search. It means they try to find exact 100% matches or not at all, so this is not useful when you want to look for choices of similar merchandise -- for example in a lighter colour or when you want to discover rather than do straightforward matching,” he said.
Tan said that image-matching is relatively easy, with ViSenze’s solution also offering that option, but when there are no exact matches for you to proceed further, it becomes a dead-end. Users then become frustrated and would need to start all over again.
According to him, the heart of ViSenze’s technology is a set of intelligent visual analytical algorithms that rapidly recognise meaningful objects in images and videos while ignoring background “noise.”
“We perform real-time recognition of objects in both images and videos, which is a key differentiator for us, and a very hard act for competitors to follow. We use adaptive machine learning and latest state-of-art computer vision techniques that allow our engine to scale quickly. We can technically train our machines to recognise almost anything under the sun,” he said enthusiastically.
When applied to fashion searches, it means the solution can automatically identify people in complex images, 'extract' their apparel and other visible fashion accessories like handbags and shoes, and use them as a basis to index and search for related or similar fashion images and products.
This technology has now been 'productised' into purpose-driven applications for e-commerce (find or discover similar items); native advertising (in-image and in-video advertising); and mobile (using camera in smartphones to take photos of fashion items like handbags and automatically have results of similar items returned to them for purchase or recommended to them for discovery).
Vision as a service
Tan is understandably proud of the founding team’s credentials, pointing out that half of them come from a science background, with the other half from a web background.
The company was founded by Yuen, a seasoned entrepreneur who serves as its non-executive chairman; Tan, who left the corporate world to serve as its full time CEO; Professor Chua Tat Seng, the current co-director of NExT who also serves as a non-executive scientific advisor to ViSenze; and Dr Li Guangda, the chief technology officer, who was the lead visual recognition researcher at NExT.
What Tan sees in this technology is the potential to disrupt existing practices and blaze a trail for a new visual search experience, as well as empower users with intelligent everyday tools that they only thought existed in R&D labs in universities.
“Who needs QR codes tomorrow when we can detect visual objects using pure visual attributes in images today?
“Like most startups with their interesting stories, our internal gag was ‘an invisible solution meets a visible problem’ -- not the other way round -- and ‘visual fusion where science meets fashion,’ because we started first by piloting heavily on the most visual of all e-commerce categories,” he added.
In terms of a business model, Tan said the company will offer its solution to digital businesses in the form of Software-as-a-Service (SaaS) or as he likes to call it, “Vision-as-as-Service.”
“We make money from volume-based usage of our visual technology, for example the number of API (applications programming interface) calls made to our visual engine hosted in the cloud. We have plans for a direct-to-consumer model that will become a direct visual search application or a tool for mobile consumers to take pictures of almost anything, and have useful and relevant information returned to them. This will be a very key product focus going forward,” he added.
The founders put in initial funding for ViSenze, and the company recently secured another US$400,000 in R&D grants from Singapore-based investors and organisations.
“We are embarking on our Series A fundraising with which we plan to increase R&D headcount, accelerate our product development and scale up our cloud platform to support global clients within the next six months. Our key markets we want to reach quickly will be North America and China,” said Tan, adding that the team is currently in discussions with some early stage investors.
ViSenze currently has eight full-time employees, including six specialists in computer vision from leading universities, and is in the process of hiring more. Growth plans involve working with key partners such as affiliate networks, ad networks and e-marketplaces in these markets. The team intends to secure these long-term projects to enable it to scale quickly to reach consumers.
When asked what has been the biggest challenge for ViSenze to date, Tan shared that breaking new ground with digital businesses that are so used to existing norms has been one hurdle.
“But I would say that we have been very fortunate in meeting many talented people who get it in five minutes, and can immediately see the potential of what we do, and how they could use us. Rakuten is one such party,” he said.
Tan said ViSenze has targeted the global market from day one -- this scaling up quickly with the right talents, resources and securing access to key markets will be the main challenges going forward.
As for his own personal challenge, he said it was needing to think out of the box every day to solve issues that have little or no references, and challenging his team to push harder on bold new ideas.
“I want to do something that I can be truly proud of before I kick the bucket. Going back to working in a startup is humbling but not something new for me as this is my second startup experience after 16 years in the corporate world. In fact, it is refreshing to flex my thinking and experiment with bold things that most big corporations wouldn’t have the appetite for,” he added.