Carnegie Mellon University

About: Carnegie Mellon University is a education organization based out in Pittsburgh, Pennsylvania, United States. It is known for research contribution in the topics: Computer science & Robot. The organization has 36317 authors who have published 104359 publications receiving 5975734 citations. The organization is also known as: CMU & Carnegie Mellon.

Understanding Convolution for Semantic Segmentation

Abstract: Recent advances in deep learning, especially deep convolutional neural networks (CNNs), have led to significant improvement over previous semantic segmentation systems. Here we show how to improve pixel-wise semantic segmentation by manipulating convolution-related operations that are of both theoretical and practical value. First, we design dense upsampling convolution (DUC) to generate pixel-level prediction, which is able to capture and decode more detailed information that is generally missing in bilinear upsampling. Second, we propose a hybrid dilated convolution (HDC) framework in the encoding phase. This framework 1) effectively enlarges the receptive fields (RF) of the network to aggregate global information; 2) alleviates what we call the "gridding issue"caused by the standard dilated convolution operation. We evaluate our approaches thoroughly on the Cityscapes dataset, and achieve a state-of-art result of 80.1% mIOU in the test set at the time of submission. We also have achieved state-of-theart overall on the KITTI road estimation benchmark and the PASCAL VOC2012 segmentation task. Our source code can be found at https://github.com/TuSimple/TuSimple-DUC.

A comparison of string distance metrics for name-matching tasks

Abstract: Using an open-source, Java toolkit of name-matching methods, we experimentally compare string distance metrics on the task of matching entity names We investigate a number of different metrics proposed by different communities, including edit-distance metrics, fast heuristic string comparators, token-based distance metrics, and hybrid methods Overall, the best-performing method is a hybrid scheme combining a TFIDF weighting scheme, which is widely used in information retrieval, with the Jaro-Winkler string-distance scheme, which was developed in the probabilistic record linkage community

Using color to separate reflection components

Abstract: In computer vision, the goal of which is to identify objects and their positions by examining images, one of the key steps is computing the surface normal of the visible surface at each point (“pixel”) in the image. Many sources of information are studied, such as outlines ofsuifaces, intensity gradients, object motion, and color. This article presents a method for analyzing a standard color image to determine the amount of interface (“specular”) and body (“diffuse”) reflection at each pixel. The interface reflection represents the highlights from the original image, and the body reflection represents the original image with highlights removed. Such intrinsic images are of interest because the geometric properties of each type of reflection are simpler than the geometric properties of intensity in a black-and-white image. The method is based upon a physical model of reflection which states that two distinct types of reflection–interface and body reflection–occur, and that each type can be decomposed into a relative spectral distribution and a geometric scale factor. This model is far more general than typical models used in computer vision and computer graphics, and includes most such models as special cases. In addition, the model does not assume a point light source or uniform illumination distribution over the scene. The properties of tristimulus integration are used to derive a new model of pixel-value color distribution, and this model is exploited in an algorithm to derive the desired quantities. Suggestions are provided for extending the model to deal with diffuse illumination and for analyzing the two components of reflection.

The mind's eye in chess.

Abstract: Publisher Summary This chapter describes the progress made toward understanding chess skill. It describes the work on perception in chess, adding some new analyses of the data. It presents a theoretical formulation to characterize how expert chess players perceive the chess board. It describes some tasks that correlate with chess skill and the cognitive processes of skilled chess players. It is believed that the demonstration of de Groot's, far from being an incidental side effect of chess skill, actually reveals one of the most important processes that underlie chess skill—the ability to perceive familiar patterns of pieces. In the first experiment discussed in the chapter, two tasks were used. The memory task was very similar to de Groot's task: chess players saw a position for 5 seconds and then attempted to recall it. Unlike de Groot, multiple trials were used—5 seconds of viewing followed by recall—until the position was recalled perfectly. The second task or the perception task for simplicity involved showing chess players a position in plain view.

Testing Intrusion detection systems: a critique of the 1998 and 1999 DARPA intrusion detection system evaluations as performed by Lincoln Laboratory

Abstract: In 1998 and again in 1999, the Lincoln Laboratory of MIT conducted a comparative evaluation of intrusion detection systems (IDSs) developed under DARPA funding. While this evaluation represents a significant and monumental undertaking, there are a number of issues associated with its design and execution that remain unsettled. Some methodologies used in the evaluation are questionable and may have biased its results. One problem is that the evaluators have published relatively little concerning some of the more critical aspects of their work, such as validation of their test data. The appropriateness of the evaluation techniques used needs further investigation. The purpose of this article is to attempt to identify the shortcomings of the Lincoln Lab effort in the hope that future efforts of this kind will be placed on a sounder footing. Some of the problems that the article points out might well be resolved if the evaluators were to publish a detailed description of their procedures and the rationale that led to their adoption, but other problems would clearly remain./par>