From Artificial Intelligence to Intelligent Imaging

Last year at Mobile World Congress I expected the buzz to be about smartphone innovations. Instead, what most people talked about was IoT, the Internet of Things. My highlight was brushing my teeth with the Oral-B Genius smart toothbrush at one of several sinks lined up in the company’s booth. The idea was: you plop your smartphone onto the mirror with a provided suction cup and the toothbrush’s position detection technology uses your smartphone’s camera as well as the brush’s motion sensors to track how well you’re brushing. You think you’re ready, but Genius disagrees? Don’t worry: it will alert you and nudge you to keep brushing. When done, it provides you with your stats, i.e. you’ll get your, uhm, brownie points.

But the IoT – that was last year.

This year, if CES was any indication, all we’ll hear about at MWC is artificial intelligence: exciting new AI-based solutions – or older ones that are repackaged as “based on AI.”

****

Hopefully, we’ll also see some exciting smart imaging innovations that leverage AI.

Why? Because we need more smartness in our photo lives: we’re taking way too many photos, which are way too difficult to keep track of, way too hard to enhance into must-keep masterpieces, and way too time-consuming to combine with other content into enticing collages, multimedia trailers, or printed photobooks.

What we need is smart imaging. And what is smart? Maybe the solution is technically based on AI (more precisely, AI’s more recently developed Deep Learning offshoot that uses complex neural networks to generate insights from massive training sets – just provide the system 100,000 photos of a ball, and it will be able to tell if there are any balls in the next photo). Or maybe it isn’t technically based on AI (your programmers describe the shape and other characteristics of a ball through a set of rules, and the next time their solution also knows if a photo includes a ball).

[source: twitter user @teenybiscuit as quoted in Mashable]

The consumer couldn’t care less about whether their smart tool uses AI or not, as long as that tool is smart, i.e. it saves them time, lets them do things they wouldn’t be able to do themselves or even suggests things that hadn’t crossed their minds.

As a user I shouldn’t have to manually tag photos that have a ball in them; my photo organizing app should do this for me – not only when instructed, but also proactively by pleasantly surprising me with an enticing photo collage of our previous holiday soccer games. And the app should not show me 250 photos to sift through, but just the 15 that it has learned that I would enjoy viewing most. Or, proactively prepare another selection with the photos grandma would prefer to see.

****

Our photo industry was a major playground for the Deep Learning algorithms when those first came out four years ago (facilitated by the large training sets of tagged photos provided by ImageNet and Flickr) and continues to develop innovative AI-based photo solutions, even if their media attention pales in comparison to the AI announcements made outside our industry.

To give a few examples (see more in The Photo Engagement Trends Report):

• Image recognition has made remarkable progress, but it is still in its early stages. The most exciting new frontier is combining the image content analysis with that of other data sources, such as metadata, social graph data, or online behavioral data. We’ll not only better understand the content of our pictures, but also the context of what is displayed.
• Whoever thought we were done developing photo enhancement apps is wrong. Last year, out of left field, a new breed of apps came out that applied art-like filters to users’ photos, using an AI-based filter technology called “style transfer.” Popularized by Prisma, apps like PicsArt, Photo Lab, Artisto, Facebook, BeFunky, and various others now also offer these types of popular filter effects for photos or videos.
• But AI is also used for other photo enhancement purposes. Camera app Microsoft Pix Camera uses AI to intelligently apply image enhancements by analyzing a total of 10 photos that the app captures in the background right before and right after the user takes their photo. Google Photos’ Looks feature uses a different approach to image enhancement: its first analyzes and understands the content of an image before it applies appropriate photo edits.
• The PhotoGurus app combines an AI-based engine with human designers to curate their users’ sprawling photo libraries and to custom-design digital and printable photobooks.
• Apps such as Magisto, Sharalike, RealTimes Moviemaker, and Storyo provide smart auto-curation and composing features, thus allowing the consumer to turn their collections of photos and/or video clips into enticing video trailers through the click of a button.

And the list of smart imaging solutions is only growing, effectively catering to each phase of the photo workflow: photo capture, organization, enhancement, combination, sharing, and printing.

While the chorus of “smart thises” and “smart thats” will only get louder in the coming months, let’s not forget the remarkable progress being made right inside our mobile photo industry. Stay tuned for what else is in store!

****

A few more things…

Shutterfly. Ouch! Shutterfly’s stock got a 15% hit after it missed its Q4 EBITDA forecast, with revenues on the low end of the outlook range. The company also announced a headcount reduction by 260, 13% of the company.

• The culprits? Disappointing results from the non-Shutterfly brands: Tiny Prints, Wedding Paper Divas, MyPublisher and BorrowLenses brands.
• To be nixed: MyPublisher, TripPix, FavPix, Pro Gallery and BorrowLenses (its oddly acquired camera rental business now for sale).
• No need to panic, as the company still raked in $208M in EBITDA for the year. But a fresh reassessment of the business + elimination of perceived dead wood is sort of what you expect when a new CEO is 8 months into the job.
• On the bright side? Mobile app customers and revenue more than doubled over the prior year. That’s why we’re really disappointed to see TripPix go: an under-marketed app that tried an innovative approach by letting users create data-rich digital and printed travel books, rather than the print-first-and-foremost approach followed by most photobook apps.

Flipagram. Video creation app Flipagram is being acquired by Toutiano, a Chinese news aggregator. Once a highflyer, Flipagram unsuccessfully and at considerable cost tried to build its own community in an effort to prevent its users from posting their videos externally on video sharing or social media sites.

Clarifai. Past Mobile Photo Connect presenter and image recognition technology company, Clarifai is nibbling at its war chest of $30M that it raised in October of last year: it opened a San Francisco office and hired former Google Brain developer Andrea Frome as the company’s head of research.

Snap(chat). And then there is the Snap IPO with lots of details now revealed in the company’s prospectus that were closely held secrets in the past. 161 million daily active users, which generate about $1 each. Still a far cry from Facebook’s 1.23 billion DAUs and a global ARPU of $4.83 (for both companies the ARPU for US users is several times higher). On average, over 25% of Snap’s DAUs post to their Snapchat Story every day. And yes, we now have hard proof of how much Instagram’s Stories impacted Snapchat’s growth. Here are all the key charts and images, and here is the prospectus.

The Business Forum Imaging speaker list is now available. Join me in Cologne, March 1-2!

Photo by Gabriele Barni