Take the latest VB Survey to share how your company is implementing AI today.

Five years after launching the first in-image ad platform, Luminate is launching what it calls a “next-generation” service which knows what’s in images, can target them by keyword or concept, and can handle both display and cost-per-click performance campaigns.

Plus, it filters out the nasty pics.

“We assume all images are guilty until proven innocent,” Luminate CEO James Everingham laughs, explaining that about 25 percent of images are not brand-safe. “Context has been the huge missing piece … algorithms have been able to identify a guy with bracelets, but the context might be someone getting arrested — not a good context for advertising.”

Example ads:


There are an almost-incomprehensible three trillion images online now, Luminate says, and we’re adding 500 million a day. That makes in-image advertising a rising tide, with competitors like GumGum, Vibrant Media, and PopMarker. Luminate says, however, that its new tech is a game-changer — and the first-ever rich media ad network specifically built for images.

VB studied mobile ad networks for our Mobile Game Monetization report.

That new tech includes better targeting of keywords of objects appearing in images, which are detected by Luminate’s object-recognition algorithms, plus a crowdsourced component that sees 20 percent of images feeding into Luminate reviewed by an actual human in a sort of Mechanical Turk method. The results are fed into a platform which allows advertisers to target keywords that match up with the images, much as they do on Google’s AdSense, or concepts such as “winning,” “happy couples,” or “families.”


Above: An example in-image ad

Perhaps even more interesting, Luminate can dynamically generate thousands or tends of thousands of ads from an online retailers’ product feed.

The results, Everingham says, are spectacular.

“The ads are native, so they’re relevant — almost editorial — and they’re out of the banner-blindness zone,” he told me. “We’ve seen 5-10X ad industry average results.”

Everingham is no stranger to tech startups. He ran the browser development group at Netscape for five years. And joining him are key hires from TellMe and LiveOps, bringing in significant text recognition and machine learning expertise. Still, he says that image recognition is the “final frontier” in computing, requiring massive intelligence and huge scale.

There’s some final frontierish about his view of the in-image ad industry, as well.

“Images are almost at the point of what I like to call search inception,” he told me. “Everything we buy we’ve seen somewhere. The normal process is: ‘Wow, that’s cool, what is it?’ … but if you’re able to satisfy that curiosity immediately, that’s magic.”

Luminate tested the platform on a campaign with McDonald’s, which wanted to target the concept of “winning” in a real-time native ad campaign. Luminate applied its technology to associate the company’s creative on images in real-time, focusing on images of scoring touchdowns, hitting home runs, and so on. For HP, the company targeted families, who are the biggest consumers of home printing and ink.

The strong results, Everingham said, are due to the fact that this is the “first truly native in-image ad format.”

One other component of the new launch is interesting: dynamic resizing for mobile.

“Images are all over the place when it comes to size, as are mobile devices,” he said. “But our system can perfectly create a really great experience on any size.”

Other customers include Best Buy, eBay, The Gap, and Intel.