Abstract

Large-scale 3D reconstruction has received a lot of attentionrecently. Bundle adjustment is a key component of the reconstructionpipeline and often its slowest and most computational resource intensive.It hasn’t been parallelized effectively so far. In this paper, we present ahybrid implementation of sparse bundle adjustment on the GPU usingCUDA, with the CPU working in parallel. The algorithm is decomposedinto smaller steps, each of which is scheduled on the GPU or the CPU. Wedevelop efficient kernels for the steps and make use of existing libraries forseveral steps. Our implementation outperforms the CPU implementationsignificantly, achieving a speedup of 30-40 times over the standard CPUimplementation for datasets with upto 500 images on an Nvidia TeslaC2050 GPU.

Abstract

Genetic algorithms are effective in solving many optimiza-tion tasks. However, the long execution time associated withit prevents its use in many domains. In this paper, we pro-pose a new approach for parallel implementation of geneticalgorithm on graphics processing units (GPUs) using CUDAprogramming model. We exploit the parallelism within achromosome in addition to the parallelism across multiplechromosomes. The use of one thread per chromosome byprevious efforts does not utilize the GPU resources effec-tively. Our approach uses multiple threads per chromosome,thereby exploiting the massively multithreaded GPU moreeffectively. This results in good utilization of GPU resourceseven at small population sizes while maintaining impressivespeed up for large population sizes. Our approach is mod-eled after the GAlib library and is adaptable to a varietyof problems. We obtain a speedup of over 1500 over theCPU on problems involving a million chromosomes. Prob-lems of such magnitude are not ordinarily attempted due tothe prohibitive computation times.

Abstract

Graphics processing units provide a large compu-tational power at a very low price which position them as anubiquitous accelerator. General purpose programming on thegraphics processing units (GPGPU) is best suited for regulardata parallel algorithms. They are not directly amenable foralgorithms which have irregular data access patterns suchas list ranking, and finding the connected components of agraph, and the like. In this work, we present a GPU-optimizedimplementation for finding the connected components of agiven graph. Our implementation tries to minimize the impactof irregularity, both at the data level and functional level.Our implementation achieves a speed up of 9 to 12 timesover the best sequential CPU implementation. For instance, ourimplementation finds connected components of a graph of 10million nodes and 60 million edges in about 500 millisecondson a GPU, given a random edge list. We also draw interestingobservations on why PRAM algorithms, such as the Shiloach-Vishkin algorithm may not be a good fit for the GPU and howthey should be modified.

Abstract

ndexing image data for content-based image search is an important area in Computer Vision. The state of theart uses the 128-dimensional SIFT as low level descriptors. Indexing even a moderate collection involves severalmillions of such vectors. The search performance depends on the quality of indexing and there is often a need tointeractively tune the process for better accuracy. In this paper, we propose a a visualization-based tool to tunethe indexing process for images and videos. We use a feature selection approach to improve the clustering of SIFTvectors. Users can visualize the quality of clusters and interactively control the importance of individual or groupsof feature dimensions easily. The results of the process can be visualized quickly and the process can be repeated.The user can use a filter or a wrapper model in our tool. We use input sampling, GPU-based processing, and visualtools to analyze correlations to provide interactivity. We present results of tuning the indexing for a few standarddatasets. A few tuning iterations result in an improvement of over 4% in the final classification performance, whichis significant.

Abstract

IISc Bangalore has developed a recognition engine for Tamil printed text, which has been tested on 1000 document images of pages scanned from books printed between 1950 and 2000. IIIT Hyderabad has developed a XML based annotated database for storing the 5000 images of scanned pages and the corresponding typed text in Unicode. CDAC, Noida has developed an efficient evaluation tool, which compares the OCR output text to the reference typed text (ground truth) and flashes the substitution, deletion and insertion errors in different colours on the screen, so that the design team can quickly identify the issues with the OCR and make corrective steps for improving the performance. IIT Delhi has proposed and developed a novel scheme for segmenting only the text regions from document images containing pictures. The OCRs uses Karhunen-Louve transform (KLT) as features and a support vector machine (SVM) classifier with RBF kernel in a discriminative directed acyclic graph (DDAG) configuration. They assume an uncompressed input image of the document page, scanned at a minimum of 300 dpi with 256 gray levels (not binary or two-level). Tamil OCR currently gives over 94% recognition accuracy at the Unicode level, evaluated on over 1000 printed pages, some of them also containing old Tamil letters.

Abstract

Prediction of electric load is very important issue for modern day power system engineers and a very good day ahead prediction of electric load is required for efficient performance of various Energy Management System (EMS) functions such as unit commitment, economic dispatch, fuel scheduling, and unit maintenance. A fuzzy based approach for day ahead prediction of electric load using Mahalanobis distance has been chosen in this work. Mahalanobis distance provides the similar characteristic days from the historical data set based on some independent variables generally of climate and time (such as temperature, day of the week, month etc.) and that are used to predict the dependent variable, i.e., day ahead electric load demand. The independent variables considered for the distance measure include the hourly humidity values, hourly temperatures values, and the day type variable.The similarity between the load on the day of prediction and that on similar characteristic days, which is evaluated using fuzzy system

Abstract

Incorporating ferromagnetic materials into integrated microwave devices is a promising approach for the development of on-chip high-performance circuit components. Therefore, high-frequency domain-wall motion and magnetization rotation, which yield permeability, are of primary interest. However, so far it has not been attempted to physically separate high-frequency domain-wall motion and magnetization rotation driven by external magnetic field excitation. Nor have attempts for the corresponding characterizations. In this work, patterned permalloy films are integrated with on-chip microstrip lines. Domain-wall motion and magnetization rotation are separated through aspect ratio and dimension control. The measured results show that high-frequency-field driven domain-wall motion is fast, different from current driven domain-wall motion. It is also shown that coupling effects are not important when the distance between two adjacent permalloy films is~ 1 μm despite their large lateral dimensions. Finally, surface topography affects domain structures and corresponding dynamic processes.

Abstract

In this paper we deduce the evolution of a four wheeled active suspension rover from a five wheeled passive suspension rover. The aim of this paper is to design a suspension mechanism which utilizes the advantages of both passive suspension and active suspension rover. Both the design considered here are simpler than the existing suspension mechanisms in the sense that the number of links as wells as the number of joints have been significantly reduced without compromising the climbing capability of the rover. We first analyze the kinematics of the five wheeled rover and its motion pattern while climbing an obstacle and try to deduce the same motion pattern and capability in the four wheeled rover. Both the suspension mechanism consists of two planar closed kinematic chains on each side of the rover. We also deduce the control strategy for the active suspension rover wherein only two actuators are used to control the internal configuration of the rover. To the best of author's knowledge this is the minimum number of actuators required to control the internal configuration of a active suspension while operating on a fully 3D rough terrain. Extensive uneven terrain simulations are performed for both 5-wheeled and 4-wheeled rover and a comparative analysis has been done on maximum coefficient of friction and torque requirements.

Abstract

The ability to detect, and track multiple moving objects like person and other robots, is an important prerequisite for mobile robots working in dynamic indoor environments. We approach this problem by detecting independently moving objects in image sequence from a monocular camera mounted on a robot. We use multi-view geometric constraints to classify a pixel as moving or static. The first constraint, we use, is the epipolar constraint which requires images of static points to lie on the corresponding epipolar lines in subsequent images. In the second constraint, we use the knowledge of the robot motion to estimate a bound in the position of image pixel along the epipolar line. This is capable of detecting moving objects followed by a moving camera in the same direction, a so-called degenerate configuration where the epipolar constraint fails. To classify the moving pixels robustly, a Bayesian framework is used to assign a probability that the pixel is stationary or dynamic based on the above geometric properties and the probabilities are updated when the pixels are tracked in subsequent images. The same framework also accounts for the error in estimation of camera motion. Successful and repeatable detection and pursuit of people and other moving objects in realtime with a monocular camera mounted on the Pioneer 3DX, in a cluttered environment confirms the efficacy of the method.

Abstract

In environments with identical features, the global localization of a robot, might result in multiple hypotheses of its location. If the situation is extrapolated to multiple robots, it results in multiple hypotheses for multiple robots. The localization is facilitated if the robots are actively guided towards locations where it can use other robots as well as obstacles to localize itself. This paper aims at presenting a learning technique for the above process of active localization of multiple robots by co-operation. An MDP framework is used for learning the task, over a semi-decentralized team of robots hereby maintaining a bounded complexity as opposed to various multi-agent learning techniques, which scale exponentially with the increase in the number of robots.