Fast, Accurate Detection of 100,000 Object Classes on a Single Machine a prizewinning paper by Google Research scientists, describes a breakthrough in machine vision that can distinguish between a huge class of objects 20,000 times faster than before.
This so-called convolution operator is one of the key operations used in computer vision and, more broadly, all of signal processing. Unfortunately, it is computationally expensive and hence researchers use it sparingly or employ exotic SIMD hardware like GPUs and FPGAs to mitigate the computational cost. We turn things on their head by showing how one can use fast table lookup — a method called hashing — to trade time for space, replacing the computationally-expensive inner loop of the convolution operator — a sequence of multiplications and additions — required for performing millions of convolutions with a single table lookup.
We demonstrate the advantages of our approach by scaling object detection from the current state of the art involving several hundred or at most a few thousand of object categories to 100,000 categories requiring what would amount to more than a million convolutions. Moreover, our demonstration was carried out on a single commodity computer requiring only a few seconds for each image. The basic technology is used in several pieces of Google infrastructure and can be applied to problems outside of computer vision such as auditory signal processing.
Fast, Accurate Detection of 100,000 Object Classes on a Single Machine
(Image: Clutter, a Creative Commons Attribution Share-Alike (2.0) image from neofob's photostream)
University of Washington profs Carl T. Bergstrom (Biology) and Jevin West (Information School) have proposed a course called “Calling Bullshit in the Age of Big Data” that characterizes “the majority of administrative activity” as “sophisticated exercise(s) in the combinatorial reassembly of bullshit” and aims to train students to “navigate the bullshit-rich modern environment by identifying […]
Isao Echizen, a researcher at Japan’s National Institute of Informatics, told a reporter from the Sankei Shimbun that he had successfully captured fingerprints from photos taken at 3m distance at sufficient resolution to recreate them and use them to fool biometric identification systems (such as fingerprint sensors that unlock mobile phones).
A paper from a group of Kings College London researchers documents an unexpected and welcome side effect from an experimental anti-Alzheimer’s drug called Tideglusib: test subjects experienced a regeneration of dentin, the bony part of teeth that sits between the pulp and the enamel.
One of the best ways to progress a career in project management is through earning recognized certifications. These certifications carry significant clout and don’t require expensive tuition or student loans. This Ultimate Project Management Certification Bundle is a great example of an affordable way to get ahead. It includes training for 9 certifications including PMP, […]
There’s nothing quite like the rush of playing against a real human opponent. But from a developer standpoint, creating fun multiplayer experiences is incredibly complex. Fortunately, the Unity3D game engine has made all aspects of game creation, including multiplayer functionality, as accessible as ever.This Unity Course Bundle introduces all of the necessary elements of creating […]
The 2016 World Series game 7 will undoubtedly go down in history as one of the greatest baseball games of all time. With endless suspense, a nefariously-timed rain delay, and extra innings, it reminded over 40 million viewers why they love America’s pastime – and why all bets were truly off in 2016. Savor the […]