software
Qualcomm releases FastCV embedded vision library for mobile devices
Qualcomm announced that is releasing FastCV, a computer vision library optimized for mobile platforms. FastCV offers hardware acceleration and an API for ARM processors. According to a Qualcomm blog post,
"FastCV is the framework at the heart of our vision-based Augmented Reality (AR) SDK, because AR is much more precise and useful when it’s based on camera input than on location-based estimates. We anticipate FastCV will be used by middleware developers to build additional frameworks that will allow developers of computer vision apps to build compelling experiences around CV that take advantage of:
* gesture recognition
* face detection, tracking and recognition
* text recognition and tracking
* depth of field calculations"
The FastCV SDK is available at: http://developer.qualcomm.com/fastcv.
Startup Magisto initiates automatic video editing service, receives $5.5 million
Computer vision startup Magisto is now making its automated video editing service publicly available. Previously available only to invitation-only members of a beta-test group, the Magisto algorithms take unedited home videos, search for the most interesting parts, and edit them into a movie. According to cnet news, "the system is designed to recognize faces -- even the most familiar faces -- as well as landscapes, animals, and movements such as zoom shots or action sequences." Magisto also received $5.5 in series B funding led by Li Ka-shing's Horizons Ventures.
Intel's River Trail boosts computer vision in web apps
Intel has developed an open-source JavaScript extension that enables web applications to exploit parallel processing capabilities of the underlying hardware. The extension, called River Trail, is currently available as a Firefox add-on at github. River Trail enables web applications to run on multiple cores and to use vector processing capabilites. The Intel software is compatible with HTML5 and with WebGL.
Mozilla CTO and JavaScript inventor Brendan Eich told The Register that "The demo shows a 15x speedup over serial JavaScript. It lightens up the ridiculously parallel hardware in modern CPUs and GPUs, for audio, video, image processing, automated voice response, computer vision, 3D gaming, etc. – all written in memory-safe, clean, functional JavaScript, without threads and their data races and deadlocks."
SentiSight 3.0 object recognition SDK released
Computer vision and biometrics company Neurotechnology has announced the release of its object recognition SDK SentiSight 3.0. Local feature based and shape based algorithms are both used for recognition. Two training modes are available: a manual mode in which objects in training images are outlined by a user, and an automatic mode in which objects are automatically detected against a static background and an object holder such as a hand. The software also includes object tracking capability.
ZigFu provides rapid-setup motion capture
Venture Beat reports that startup Zigfu is releasing software to enable developers to set up motion capture in minutes. The software integrates the OpenNI framework for natural gestural interaction, the Unity3D game engine, and the PrimeSense NITE middleware for 3D perception. According to ZigFu co-founder Amir Hirsch, "It takes about two minutes of click-and-drag work to get your own motion-controlled avatar up and running. The task involves connecting the shoulders, elbows, knees and hips on your 3-D character to the data output by the skeleton tracking module." The software also supports the Kinect SDK.
Related articles:
Layar Vision augmented reality toolkit released, $55,000 contest announced
Layar, maker of a platform enabling development of augmented reality (AR) layers for the real world, has announced the availability of Layar Vision, a computer vision toolkit for the Layar system. Layar Vision enables recognition of up to 50 targets to be programmed as part of a real-world user interace. In conjunction with the release of Layar Vision, Layar announced a contest to award $55,000 in prizes to 10 developers of new layers that incorporate Layar Vision. Layar Vision is available for both Android and iOS. More information is available in a ReadWriteWeb.com article.
Microsoft bundles Kinect SDK with Robotics Developer Studio
According to EE Times, Microsoft has bundled its Kinect software development kit (SDK) with its free Robotics Developer Studio. The SDK would enable access to Kinect's raw data, as well as its gesture-recoginition algorithms. The article also states that Microsoft plans to add direct support for autonomous navigation tasks. Although developers are not allowed to use the Kinect SDK to develop commercial products, nonprofit organizations will be able to add navigational algorithms for maneuvering robots.
Open-source server software for object tracking on Kinect
Open-source software (OSS) for tracking objects in 3D using depth information from a Kinect has been made available at https://code.google.com/p/intrael/. The tracking software runs on a server and communicates with a local client via HTTP; the results can be integrated into a web page using Javascript. The ability to track 24+ objects simultaneously is claimed; however, it appears to be a blob-based tracker, so that the objects must be separated from each other in space to be separately tracked.
Embedded Vision Alliance launches
Press release from Embedded Vision Alliance
More than fifteen leading technology companies — including some of the largest semiconductor companies — have joined forces to speed the adoption of computer vision capabilities in electronic products. The ability of machines to see and understand their environments — what we call “embedded vision” — promises to transform the electronics industry with products that are more intelligent and aware of their environments, and to create significant new markets for electronic equipment and components. A new consortium, called the Embedded Vision Alliance (www.embedded-vision.com), will enable the proliferation of embedded vision technology by providing design engineers with information, practical know-how, and industry standards.
“Adding computer vision to embedded systems creates phenomenal new products, markets, and opportunities,” according to Jeff Bier, president of BDTI. “Just look at the Microsoft KinectTM, which added vision to the Xbox 360TM—it became the fastest-selling consumer electronics device in history, shipping more than 10 million units in 5 months. But that’s just a small part of the story. From automobiles that prevent accidents to security cameras that prevent crimes, embedded vision will proliferate across a multitude of markets.”
MATLAB releases new Computer Vision System Toolbox
MathWorks has released a Computer Vision System Toolbox in its MATLAB version 4.0 (R2011a). This toolbox replaces the Video and Image Processing Blockset, and includes additional computer vision algorithms for feature extraction, feature matching, uncalibrated stereo rectification, and many other capabilities. More information is available in a Mathworks blog by Steve Eddins, a software development manager in the MATLAB and image processing areas at Mathworks.