Another acquisition in the area of image recognition technology, and another exit for a European startup to a U.S. giant. Qualcomm has acquired Euvision Technologies, a specialist in image recognition applications powered by artificial intelligence, originally spun out of the University of Amsterdam in The Netherlands. Its first publicly released app was Impala — an app for iOS and Android that essentially “reads” and organises photos on your phone into different categories (we reviewed it here) — without needing to work in the cloud. That app, at the time of writing, still appears to be live.
Neither Euvision nor Qualcomm itself have made any announcement of the acquisition directly, although the latter company’s investment arm, Qualcomm Ventures, has tweeted the news (Euvision was a 2014 finalist in Qualcomm Ventures’ seed funding competition, the QPrize). The acquisition is reportedly worth “tens of millions” according to the Dutch blog Startup Juncture.
We are trying to get a confirmation of a price, as well as more detail on what Qualcomm plans to do with the new assets, and will update this as we learn more.
There are a couple of reasons why the acquisition of Euvision makes sense for Qualcomm.
Euvision is Qualcomm’s 22nd acquisition, by CrunchBase’s count, and the second that Qualcomm has made this year in the area of image recognition. In January, it acquired Kooaba, a Swiss startup, whose tech has been integrated into Qualcomm’s Vuforia visual recognition service. In other words, there is already an existing business that Qualcomm is building up, where Euvision could fit.
For its part, Euvision notes that while its technology can be used for fun, consumer applications like Impala, the bigger vision is to use it for much larger and heavier enterprise-style purposes — for example, mining for images for large businesses like social networks or organizations like Interpol.
This very much taps into bigger themes we’ve seen around whether and how popular sites like Twitter, for example, cope with one current rock and hard place: making inline images appear automatically for users to drive engagement, but also blocking pictures that may be NSFW, violent or graphic in other ways.
Euvision explains its dogfooded background like this:
“As a spin-off company from the University of Amsterdam, we invented our own computer vision algorithms, scanning each and every scientific publication in our field along the way. If it was good, we absorbed it, improved it, and integrated it into our Impala software code base…”
Prior to the acquisition, Euvision had been working not just on categorization and tagging but also image mining techology that works on images never viewed before (therefore less reliant on pictures to “train” the system to look for something specific). This is/was due to come out in a few months.
“These are the first software applications that can see beyond license plates and face recognition (yawn),” the company notes. “Our software recognizes twelve different types of child pornography. It does not depend on the Interpol database, it also works for new never seen before material. And it recognizes ID cards, checks, Excel screenshots, sunsets and beaches, cats and dogs, and much more. Our software saves you time, resources and lots of frustration. And it gets your user generated content website off the Adsense black list.”
There is already some common ground between the two companies. In addition to being a finalist in the QPrize, Euvision built Impala in part on Qualcomm’s MARE (Multicore Asynchronous Runtime Environment)
computing technology. This helps speed up how Impala works on multicore phones, and also makes programming for such environments easier.
With the rise of camera-equipped smartphones, we’ve seen an explosion in the number of photos being taken and shared among people, but also a push to make editing and categorizing those images more automatic. Image recognition has been a hot area for those reasons, with companies like Facebook and Google also buying up technology and talent in this space to build out their products.
Comment