80 Million Tiny Images

Antonio Torralba (MIT), Rob Fergus (NYU), William T. Freeman (MIT)

With the advent of the Internet, billions of images are now freely available online and constitute a dense sampling of the visual world. Using a variety of non-parametric methods, we explore this world with the aid of a large dataset of 79,302,017 images collected from the Internet.


Removing Camera Shake from a Single Image

Rob Fergus (NYU), Barun Singh (MIT), Aaron Hertzmann (Toronto), Sam T. Roweis (Toronto), William T. Freeman (MIT)

Camera shake during exposure leads to objectionable image blur and ruins many photographs. Conventional blind deconvolution methods typically assume frequency-domain constraints on images, or overly simplified parametric forms for the motion path during camera shake. Real camera motions can follow convoluted paths, and a spatial domain prior can better maintain visually salient image characteristics. We introduce a method to remove the effects of camera shake from seriously blurred images. The method assumes a uniform camera blur over the image and negligible in-plane camera rotation.


GreenDot: Body-Signature Analysis

Chris Bregler (NYU), Peggy Hackney (UC Berkeley), Ian McDowall (Fakespace), Sally Rosenthal, George Williams (NYU)

GreenDot is a research project that investigates vision and machine learning techniques to detect human body language in video. The goal of the project is to train a computer to recognize a person based on his or her motions, and to identify the person's emotional state, cultural background, and other attributes. The research is conducted by an interdisciplinary team of computer scientists, movement experts, linguists, and other specialists. The current focus is the analysis of national and international public figures while they are giving speeches, with future plans to investigate many other domains. The research team is building a large database of people's motions, using TV recordings and web video downloads.