We demonstrate the first large-scale image browsing system applied to 200,000 popular queries which utilizes image content to organize image search results. Given a query, the system extracts image content features such as color, shape, local features, face signatures and metadata from up to 1000 image results, and hierarchically clusters them to form an exemplar tree. A dynamic web-based user interface allows the user to navigate this hierarchy, allowing fast and interactive browsing. The exemplars of each cluster provide a comprehensive visual overview of the query results, and allow the user to quickly navigate to the images of interest.

The proliferation of the Web and digital photography have made large scale image collections containing billions of images a reality. Image collections on this scale make performing even the most common and simple computer vision, image processing, and machine learning tasks nontrivial. An example is nearest neighbor search, which not only serves as a fundamental subproblem in many more sophisticated algorithms, but also has direct applications, such as image retrieval and image clustering. In this paper, we address the nearest neighbor problem as the first step towards scalable image processing. We describe a scalable version of an approximate nearest neighbor search algorithm and discuss how it can be used to find near duplicates among over a billion images

We present a Bayesian approach to color constancy which utilizes a non- Gaussian probabilistic model of the image formation process. The pa- rameters of this model are estimated directly from an uncalibrated image set and a small number of additional algorithmic parameters are chosen using cross validation. The algorithm is empirically shown to exhibit RMS error lower than other color constancy algorithms based on the Lambertian surface reflectance model when estimating the illuminants of a set of test images. This is demonstrated via a direct performance comparison utilizing a publicly available set of real world test images and code base.

Appearance based object detection systems utilizing statistical models to capture real world variations in appearance have been shown to exhibit good detection performance. The parameters of these statistical models are typically learned automatically from labeled training images. This process can be difficult in that a large number of labeled training examples may be needed to accurately model appearance variation. In this work we describe a method whereby a training set consisting of a small number of fully labeled training examples augmented with a set of weakly labeled examples can be used to train a detector which exhibits performance better than that which can be obtained with a reduced set of fully labeled training examples alone.

In this paper we describe a functioning low cost embedded vision system which can perform basic color blob tracking at 16.7 frames per second. This system utilizes a low cost CMOS color camera module and all image data is processed by a high speed, low cost microcontroller. This eliminates the need for a separate frame grabber and high speed host computer typically found in traditional vision systems. The resulting embedded system makes it possible to utilize simple color vision algorithms in applications like small mobile robotics where a traditional vision system would not be practical.

This paper considers a specific type of interaction: short-term and spontaneous interaction with crowds of people. Such patterns of interactions are found when service robots operate in public places, for example information kiosks, receptionists, and tour-guide robots applications. We describe our approach to spontaneous short-term interaction: a robot designed to be a believable social agent. The approach has been implemented using a mobile robot with a motorized face as focal point for interaction, an architecture that suggests the robot has moods, and a method for learning how to interact with people. Our system was recently deployed at a Smithsonian museum in Washington, DC. During a two week period it interacted with thousands of people. The robot's interactive capabilities were essential for its high on-task performance, and thus its practical success.

This paper describes an interactive tour-guide robot, which was successfully exhibited in a Smithsonian museum. During its two weeks of operation, the robot interacted with thousands of people, traversing more than 44 km at speeds of up to 163 cm/sec. Our approach specifically addresses issues such as safe navigation in unmodified and dynamic environments, and short-term human-robot interaction. It uses learning pervasively at all levels of the software architecture.

The planar thallium-201 (201Tl) myocardial perfusion scintigram is a widely used diagnostic technique for detecting and estimating the risk of coronary artery disease. Interpretation is currently based on visual scoring of myocardial defects combined with image quantitation and is known to have a significant subjective component. Neural networks learned to interpret thallium scintigrams as determined by both individual and multiple (consensus) expert ratings. Four different types of networks were explored: single-layer, two-layer backpropagation (BP), BP with weight smoothing, and two-layer radial basis function (RBF). The RBF network was found to yield the best performance (94.8% generalization by region) and compares favorably with human experts. We conclude that this network is a valuable clinical tool that can be used as a reference "diagnostic support system" to help reduce inter- and intraobserver variability. This system is now being further developed to include other variables that are expected to improve the final clinical diagnosis.