How Computer Vision Is Transforming MediaComputer Vision
HAI School: Aimed at Education Market, Google AIY Kits Coming to Target9to5Google
Google just made creating your own computer vision model to recognize, say, cats or the Empire State Building or a Malibu sunset, a little easier. The company has released a companion mobile app for its two DIY AI bundles that allows you to connect more easily with the kits’ Raspberry Pi Google Cardboard computer and the cloud when creating your models. Aiming at the education market, the Google AIY Vision and Voice Kits will also be sold in Target stores, inevitably making high school robotics teams the new hotbeds of innovation.
Studio 50: Projection Mapping on Disco Balls Creates Rave-Ready Light ShowColossal
Projection mapping, a technology that uses computer vision to dynamically project light onto real-world objects, even moving ones, in a realistic way, is already being used to transform the looks of stadiums, buildings and even t-shirts, kind of like Animojis for physical objects. But when that physical object is a disco glitterball, the effect is magnified thanks to multiple reflections. That’s precisely what digital artist-cum-coders Kyle McDonald and Jonas Jonegan pulled off in Light Leaks,an installation currently at the Los Angeles Music Center that features light projection-mapped precisely onto a group of 50 disco balls and other surfaces, which creates further reflective effects inside the room. The overall experience is nothing short of rave-tastic, but next time they should round up the number of disco balls to 54.
Nike Buys Computer Vision Startup to Scan Shoppers’ Feet for Better-Fitting ShoesFootwear News
In a bid, no doubt, to lessen the number of online returns of shoes that don’t fit, Nike last week announced the acquisition of Tel Aviv-based computer vision company Invertex, which has a platform that mixes AI and 3D mapping to scan shoppers’ feet and then recommends specific models and sizes. The company has in-store and at-home solutions, the latter of which is as easy as pointing your smartphone at your feet to get a scan, which you can then use to order optimized-fit shoes online. Running to the store for the perfect pair of shoes will be a thing of the past, enabling you to more easily just do it.
Yabba Dabba Deep Learning: AI Auto-Generates Flintstones ScenesEngadget
The Flintstones was never complex material, both in terms of writing or animation—just some Honeymooners with Stone Age-era puns and basic animations on top of still backgrounds—so it’s no surprise that a group of computer scientists are trying to get AI to recreate the show. The University of Illinois researchers have developed CRAFT (Composition, Retrieval, and Fusion Network), an AI system that was trained on 25,000 three-second clips from the 1960s animated series. Each clip was annotated with descriptions of characters, objects, and what’s going on in the scene. CRAFT was then able to edit pull together characters, objects and backgrounds automatically just by analyzing a line or two. According to the researchers, the system is apparently also able to do this from non-annotated clips, though it’s still a work in progress. Some of the resulting cartoons are not quite ready for primetime Saturday morning—especially those generated out of long and complex sentences—but CRAFT is a fascinating glimpse at how some of the more repetitive animation tasks might be automated in the coming years. This will give animators more time to really knock big original scenes out the park, or at least that’s the theory, but for now, it might be good for cartoon creatives to see if Studio Ghibli is hiring.