iStock // nikolay100
iStock // nikolay100

Watch How Computers Perform Optical Character Recognition

iStock // nikolay100
iStock // nikolay100

Optical Character Recognition (OCR) is the key technology in scanning books, signs, and all other real-world texts into digital form. OCR is all about identifying a picture of written language (or set of letters, numbers, glyphs, you name it) and sorting out what specific characters are in there.

OCR is a hard computer science problem, though you wouldn't know it from its current pervasive presence in consumer software. Today, you can point a smartphone at a document, or a sign in a national park, and instantly get a pretty accurate OCR read-out...and even a translation. It has taken decades of research to reach this point.

Beyond the obvious problems—telling a lowercase "L" apart from the number "1," for instance—there are deep problems associated with OCR. For one thing, the system needs to figure out what font is in use. For another, it needs to sort out what language the writing is in, as that will radically affect the set of characters it can expect to see together. This gets especially weird when a single photo contains multiple fonts and languages. Fortunately, computer scientists are awesome.

In this Computerphile video, Professor Steve Simske (University of Nottingham) walks us through some of the key computer science challenges involved with OCR, showing common solutions by drawing them out on paper. Tune in and learn how this impressive technology really works:

A somewhat related challenge, also featuring Simske, is "security printing" and "crazy text." Check out this Computerphile video examining those computer science problems, for another peek into how computers see (and generate) text and imagery.

nextArticle.image_alt|e
iStock
A New App Interprets Sign Language for the Amazon Echo
iStock
iStock

The convenience of the Amazon Echo smart speaker only goes so far. Without any sort of visual interface, the voice-activated home assistant isn't very useful for deaf people—Alexa only understands three languages, none of which are American Sign Language. But Fast Company reports that one programmer has invented an ingenious system that allows the Echo to communicate visually.

Abhishek Singh's new artificial intelligence app acts as an interpreter between deaf people and Alexa. For it to work, users must sign at a web cam that's connected to a computer. The app translates the ASL signs from the webcam into text and reads it aloud for Alexa to hear. When Alexa talks back, the app generates a text version of the response for the user to read.

Singh had to teach his system ASL himself by signing various words at his web cam repeatedly. Working within the machine-learning platform Tensorflow, the AI program eventually collected enough data to recognize the meaning of certain gestures automatically.

While Amazon does have two smart home devices with screens—the Echo Show and Echo Spot—for now, Singh's app is one of the best options out there for signers using voice assistants that don't have visual components. He plans to make the code open-source and share his full methodology in order to make it accessible to as many people as possible.

Watch his demo in the video below.

[h/t Fast Company]

nextArticle.image_alt|e
Ralph Gatti, AFP/Getty Images
The 'David Bowie Is' Exhibition Is Coming to Your Smartphone
 Ralph Gatti, AFP/Getty Images
Ralph Gatti, AFP/Getty Images

"David Bowie is," an exhibition dedicated to the life, work, and legacy of the pop icon, concluded its six-year world tour on July 15. If you didn't get a chance to see it in person at its final stop at New York City's Brooklyn Museum, you can still experience the exhibit at home. As engadget reports, the artifacts displayed in the collection will be recreated in virtual and augmented reality.

The Victoria and Albert Museum, the curator of the exhibit, and the David Bowie Archive are collaborating with Sony Music Entertainment and the sound and media studio Planeta on the new project, "David Bowie is Virtual." Like the physical exhibition, the digital experience will integrate visual scenes with the music of David Bowie: 3D scans will bring the musician's costumes and personal items into the virtual sphere, allowing viewers to examine them up close, and possibly in the case of the outfits, try them on.

"These new digital versions of ‘David Bowie is’ will add unprecedented depth and intimacy to the exhibition experience, allowing the viewer to engage with the work of one of the world’s most popular and influential artists as never before," the announcement of the project reads. "Both the visual richness of this show and the visionary nature of Bowie and his art makes this a particularly ideal candidate for a VR/AR adaptation."

"David Bowie is Virtual" will be released for smartphones and all major VR and AR platforms sometimes this fall. Like the museum exhibition, it will come with an admission price, with a portion of the proceeds going toward the Victoria and Albert Museum and the Brooklyn Museum.

[h/t engadget]

SECTIONS

arrow
LIVE SMARTER
More from mental floss studios