Abstract

Road networks are the lifeline of a city and understanding its usage has a number of potential applications from transportation planning and engineering aspects to environmental management. While the full network is important to be analyzed for such applications, prudent planning needs one to identify the significant sections of the road network and prioritize them. Annual Average Daily Traffic (AADT), an estimate of the average daily traffic along a defined road segment, is one such data that helps in such endeavors. But, roads are also about connectivity and accessibility across different regions. Hence, this paper proposes a study that integrates the AADT data with implicit information derived from the road network to generate a computationally representative (CoRe) well connected sub-network, significantly smaller than the original network. While the AADT data analysis looks for road segments with high traffic, this paper proposes and evaluates a graph theory based approach for calculating road priorities purely based on the topological structure of the road network. The work further demonstrates the utility of the CoRe sub-network in terms of both, achieving gains in path computation and capturing the behavioral pattern of travelers. A case study of the Melbourne, Australia supports the feasibility and applicability of this knowledge integration approach.

Abstract

In this paper, a novel technique is presented to detect buildings from very high resolution satellite image. This work builds on the learning of ICA based building detection technique from the very high resolution (VHR) multispectral satellite images presented in [1]. The candidate building pixels obtained through ICA are used to extract maximally stable extremal regions (MSER) which are then filtered using geometric properties to obtain final potential buildings. The technique is aimed at reducing false detection at pixel-level and improving object-level performance of [1]. Combining the two works offers an unsupervised building detection technique which is robust towards size, shape, color, types of rooftops and shadows. A wider test image set consisting of 15 images of different dimensions are used to evaluate performance of the complete detection process. The combined technique achieves object-level precision and recall of 80.64% and 83.65% respectively.

Abstract

In many cities, it is difficult for authorities to build road monitoring systems that are cost-friendly and effective. The existing road monitoring systems include heavy machinery with sensors fitted which make them expensive for authorities to use for frequent monitoring. To address this challenge, we propose an innovative collaborative system for road condition monitoring leveraging the existing technologies by data fusion approach. The mobile application on smart-phones of geo-citizens coordinates the data collection and reports back to a central server for assimilation and further processing. Image data is processed to obtain the road surface while the accelerometer and GPS data is used for identifying road condition and location information. This data is then fused based on location and timestamp information and stored as segments in the database. As the data obtained is crowd-sourced, there are multiple tracks for the same segment of the road and this segmented approach allows for collating multiple user-provided information and validation. In case there is multiple track information available for a particular segment, a ranking model based on the number of collaborators is used to rank the status. The final database holds the information like segmentId, timestamp, GPS coordinates and status of the segment. This data serves as an input to the visualization service. The multiplicity of data points and anonymity of the users establishes robustness and crowd-sourced model ensures frequent and continuous coverage.

Abstract

Static power consumption is one of the most critical issues in CMOS digital circuits, and FinFET technology is being recognized as a valid solution for the problem. In this chapter, we utilize a logic-level leakage current estimation technique relying on an internal node voltage-based model. The model is implemented in the form of VHDL packages. By utilizing the capability of the model, the behavior of major leakage component has been analyzed separately for FinFET technology scaling over single- and multi-stage digital standard cells.

Abstract

The paper considers direction of arrival (DOA) estimation from long-term observations in a noisy environment. In such an environment the noise source might evolve, causing the stationary models to fail. Therefore a heteroscedastic Gaussian noise model is introduced where the variance can vary across observations and sensors. The source amplitudes are assumed independent zero-mean complex Gaussian distributed with unknown variances (i.e. the source powers), leading to stochastic maximum likelihood (ML) DOA estimation. The DOAs of plane waves are estimated from multi-snapshot sensor array data using sparse Bayesian learning (SBL) where the noise is estimated across both sensors and snapshots. Simulations demonstrate that taking the heteroscedastic noise into account improves DOA estimation.

Abstract

Post-capture control of the focus position of an image is a useful photographic tool. Changing the focus of a single image involves the complex task of simultaneously estimating the radiance and the defocus radius of all scene points. We introduce RefocusGAN, a deblurthen-reblur approach to single image refocusing. We train conditional adversarial networks for deblurring and refocusing using wide-aperture images created from light-fields. By appropriately conditioning our networks with a focus measure, an in-focus image and a refocus control parameter δ, we are able to achieve generic free-form refocusing over a single image.

Abstract

Intrinsic Image Decomposition (IID) is a challenging and interesting computer vision problem with various applications in several fields. We present novel semantic priors and an integrated approach for single image IID that involves analyzing image at three hierarchical context levels. Local context priors capture scene properties at each pixel within a small neighborhood. Mid-level context priors encode object level semantics. Global context priors establish correspondences at the scene level. Our semantic priors are designed on both fixed and flexible regions, using selective search method and Convolutional Neural Network features. Experiments and analysis of our method indicate the utility of our weak semantic priors and structured hierarchical analysis in an IID framework. We compare our method with the current state-of-the-art and show results with lesser artifacts. Finally, we highlight that proper choice and encoding of prior knowledge can produce competitive results compared to end-to-end deep learning IID methods, signifying the importance of such priors. We believe that the insights and techniques presented in this paper would be useful in the future IID research.

Abstract

Replacing overexposed or dull skies in outdoor photographs is a desirable photo manipulation. It is often necessary to color correct the foreground after replacement to make it consistent with the new sky. Methods have been proposed to automate the process of sky replacement and color correction. However, many times a color correction is unwanted by the artist or may produce unrealistic results. We propose a data-driven approach to sky-replacement that avoids color correction by finding a diverse set of skies that are consistent in color and natural illumination with the query image foreground. Our database consists of ∼ 1200 natural images spanning many outdoor categories. Given a query image, we retrieve the most consistent images from the database according to L2 similarity in feature space and produce candidate composites. The candidates are re-ranked based on realism and diversity. We used pre-trained CNN features and a rich set of hand-crafted features that encode color statistics, structural layout, and natural illumination statistics, but observed color statistics to be the most effective for this task. We share our findings on feature selection and show qualitative results and a user-study based evaluation to show the effectiveness of the proposed method.

Abstract

t. In addition to maximizing the revenue, retailers also aim at diversifying product offerings for facilitating sustainable revenue generation in the long run. Thus, it becomes a necessity for retailers to place appropriate itemsets in a limited k number of premium slots in retail stores for achieving the goals of revenue maximization and itemset diversification. In this regard, research efforts are being made to extract itemsets with high utility for maximizing the revenue, but they do not consider itemset diversification i.e., there could be duplicate (repetitive) items in the selected top-utility itemsets. Furthermore, given utility and support thresholds, the number of candidate itemsets of all sizes generated by existing utility mining approaches typically explodes. This leads to issues of memory and itemset retrieval times. In this paper, we present a framework and schemes for efficiently retrieving the top-utility itemsets of any given itemset size based on both revenue as well as the degree of diversification. Here, higher degree of diversification implies less duplicate items in the selected top-utility itemsets. The proposed schemes are based on efficiently determining and indexing the topλ high-utility and diversified itemsets. Experiments with a real dataset show the overall effectiveness and scalability of the proposed schemes in terms of execution time, revenue and degree of diversification w.r.t. a recent existing scheme.

Abstract

Deep convolutional features for word images and textual embedding schemes have shown great success in word spotting. In this work, we follow these motivations to propose an End2End embedding framework which jointly learns both the text and image embeddings using state of the art deep convolutional architectures. The three major contributions of this work are: (i) an End2End embedding scheme to learn a common representation for word images and its labels, (ii) building a state of art word image descriptor and demonstrating its utility as off-the-shelf features for word spotting, and (iii) use of synthetic data as a complementary modality to further enhance word spotting and recognition. On the challenging IAM handwritten dataset, we report a mAP of 0.9509 for query-by-string based retrieval task. Under lexicon based word recognition, our proposed method report a 2.66 and 5.10 CER and WER respectively.