A group of researchers at Adobe, Microsoft, and MIT have taken normal objects and turned them into visual microphones. They discovered they can reproduce audio from silent video recordings by studying the movement of objects affected by sound waves traveling through the air.
Computer Vision Blog
The increase in accuracy in image recognition was seen at this year’s ImageNet Large Scale Visual Recognition Challenge (ILSVRC2014). The academic contest has been run annually since 2010, this year attracting thirty eight participants from thirteen different countries.
Never Ending Image Learner (NEIL) is a computer program that works 24/7 learning information about images that it finds on the internet. NEIL, which is housed at Carnegie Mellon University, is not looking for just any type of information. Rather, it’s goal is to learn common sense relationships found in everyday life.
Digital cameras have long been trying to replicate what is seen by the human eye. Now technology is advancing into what can be seen by a bug’s eye. These new devices are aimed to be capable of achieving a panoramic view with a sharp focus that can be seen at any distance.
TUTORIAL: Dead simple object tracking tutorial with line-by-line code explanations using Python, OpenCV, and the CamShift algorithm.
The Goal-Line system can track the motion of the ball up to 500 images per second. The supporting computer vision software tracks the movement of all objects on the pitch in all the images and filters out the players, referees and all disturbing objects.
Chris Kluwe wants to look into the future of sports and think about how technology will help not just players and coaches, but fans. Here the former NFL punter envisions a future in which augmented reality will help people experience sports as if they are directly on the field — and maybe even help them see others in a new light, too.
The Facebook team adopted Deep Learning to apply to their face verification algorithm in lieu of well engineered features which is common in majority of contributions in this field.
For image processing and computer vision applications, the Wolfram Language provides built-in support for both programmatic and interactive modern industrial-strength image processing—fully integrated with the Wolfram Language's powerful mathematical and algorithmic capabilities.
Primarily, the system works through the aid of six motion-capture cameras equipped with computer vision throughout the arena. These six synchronized cameras were divided into three per half-court areas. The cameras are programmed to take 25 images per second and record them in its memory for review by sports analytics.