Search

Deep neural networks are a key component of behavior prediction and motion generation for self-driving cars. One of their main drawbacks is a lack of transparency: they should provide easy to interpret rationales for what triggers certain behaviors. We propose an architecture called Attentional Bottleneck with the goal of improving transparency. Our key idea is to combine visual attention, which identifies what aspects of the input the model is using, with an information bottleneck that enables the model to only use aspects of the input which are important. This not only provides sparse and interpretable attention maps (e.g. focusing only on specific vehicles in the scene), but it adds this transparency at no cost to model accuracy. In fact, we find improvements in accuracy when applying Attentional Bottleneck to the ChauffeurNet model, whereas we find that the accuracy deteriorates with a traditional visual attention model.

The aim of this study is to design and develop a blockchain-based web app called Health Stack to maintain accurate and complete medical records of patients, to help doctors to fetch previous medical history of the patients, to assist user to find out the disease he or she is suffering from and much more. For medical services, secure data storage is one of major concern for people. This problem can be resolved by developing an app using a blockchain technology having the features of decentralization and verifiability. Development of this app doesn’t involve any kind of dependency on third-party.

Deep neural networks are a key component of behavior prediction and motion generation for self-driving cars. One of their main drawbacks is a lack of transparency: they should provide easy to interpret rationales for what triggers certain behaviors. We propose an architecture called Attentional Bottleneck with the goal of improving transparency. Our key idea is to combine visual attention, which identifies what aspects of the input the model is using, with an information bottleneck that enables the model to only use aspects of the input which are important. This not only provides sparse and interpretable attention maps (e.g. focusing only on specific vehicles in the scene), but it adds this transparency at no cost to model accuracy. In fact, we find slight improvements in accuracy when applying Attentional Bottleneck to the ChauffeurNet model, whereas we find that the accuracy deteriorates with a traditional visual attention model.

With complex applications and system requirements, the level of integration in modern day SoCs have increased many folds. Analog to digital Converters (ADCs) are important part of many networking and communication chips. Higher data rate requirements need ADC that can operate at high speed without much compromise in performance. As the speed, complexity and number of Analog to Digital Converters increases in a SoC, it makes its characterization and validation a challenging task. Though the tests that need to be performed are well defined in theory, performing those tests accurately to evaluate the system’s true performance is vital. The results obtained may not be as accurate as they seem, resulting in underestimating the product specifications. This paper focuses on pitfalls of dynamic measurements and propose methods to overcome them for true and accurate measurements of the Device Under Test (DUT). We also propose efficient and robust algorithms which can ease the process of calibration for more refined characterization.

"Blind race ends in a pit". A similar scenario is observed with the use of conventional dosage forms for different pathological conditions. Of late various disease states had regularly been reported to bear direct concurrence to the body's secretions that bear a constant rotationary cycle. Moreover the pharmacokinetic as well as pharmacodynamic responsiveness of various drugs had been reported to bear constant swings with the changing hours of the day. Thus, usage of the conventional zero order dosage form for every disease state or every active moiety developed, will only leave the researchers as well as the consumers in doldrums. Chronomodulated dosage forms are a silver lining in these overshadowed clouds. They are the dosage forms that spearhead the innovative researches because of their pre-programmed and pre-regulated pulsed release of the drug, at desired sites. Drug release pattern from these dosage forms considerably mimics the circadian timing of the body's secretion which are held responsible for the symptoms of the pathological irregularities arising in one's body. Thus, in a way these dosage forms are a shield against the inducers of the disease symptoms. Current review enlists the pathological states for which the chronomodulated delivery systems can prove to be a miraculous cure. This review emphasizes to summarize the patents granted as well as the novel researches undertaken by various researchers to upgrade the previously existing dosage form scenario. Moreover, this work is an attempt to summarize the various proprietary techniques and marketed formulations, thus trying to help the researchers to fabricate a better and novel dosage form from previously existing ones.

The problem of training a classifier from a handful of positive examples, without having to supply class specific negatives is of great practical importance. The proposed approach to solving this problem builds on the idea of training LDA classifiers using only class specific foreground images and a large collection of unlabelled images, as described in [11]. While we adopt the LDA training methodology of [11], we depart from HOG features and work with those extracted from a Convolutional Neural Network (CNN) pre-trained on Image Net (Over feat). We combine Over feat features with the LDA training methodology to derive generative classifiers. When evaluated on a K-way classification problem, these classifiers are almost as good as those trained discriminatively using the same features. Unlike the HOG based approach of [11], our classifiers do not need any post-processing step of calibration, a step that requires positives and negatives. Finally, we show that in an instance retrieval setup, we can employ these generative classifiers to derive a novel query-expansion framework that achieves a significant performance boost by utilizing only the top ranked positive examples from an initial nearest-neighbor list.

We propose a simple yet effective multi-feature fusion approach based on regression models for logo retrieval. Rather than fusing original features, we focus on similarities between pairs of images from multiple features, where only an annotation of similar/dissimilar pairs of images is needed. For each pair of images, a new vector is constructed by concatenating the similarities between the image pair from multiple features. A regression model is fitted on the new set of vectors with similar/dissimilar annotations as labels. Similarities from multiple features between the query and database images can then be converted to a new similarity score using the learned regression model. Initially retrieved database images are then re-ranked using the similarities predicted by the regression model. Logo class information from the training samples can also be included in the training process by learning an ensemble of regression models for individual logo classes. Extensive experiments on public logo datasets Flickrl.ogol Z and Belga Logo demonstrate the effectiveness and superior generalization ability of our approach for fusing various features.

The sheer growth in data volume and Hadoop cluster size make it a significant challenge to diagnose and locate problems in a production-level cluster environment efficiently and within a short period of time. Often times, the distributed monitoring systems are not capable of detecting a problem well in advance when a large-scale Hadoop cluster starts to deteriorate i n performance or becomes unavailable. Thus, inc o m i n g workloads, scheduled between the time when cluster starts to deteriorate and the time when the problem is identified, suffer from longer execution times. As a result, both reliability and throughput of the cluster reduce significantly. In this paper, we address this problem by proposing a system called Astro, which consists of a predictive model and an extension to the Hadoop scheduler. The predictive model in Astro takes into account a rich set of cluster behavioral information that are collected by monitoring processes and model them using machine learning algorithms to predict future behavior of the cluster. The Astro predictive model detects anomalies in the cluster and also identifies a ranked set of metrics that have contributed the most towards the problem. The Astro scheduler uses the prediction outcome and the list of metrics to decide whether it needs to move and reduce workloads from the problematic cluster nodes or to prevent additional workload allocations to them, in order to improve both throughput and reliability of the cluster. The results demonstrate that the Astro scheduler improves usage of cluster compute resources significantly by 64.23% compared to traditional Hadoop. Furthermore, the runtime of the benchmark application reduced by 26.68% during the time of anomaly, thus improving the cluster throughput.

We propose a purely geometric correspondence-free approach to urban geo-localization using 3D point-ray features extracted from the Digital Elevation Map of an urban environment. We derive a novel formulation for estimating the camera pose locus using 3D-to-2D correspondence of a single point and a single direction alone. We show how this allows us to compute putative correspondences between building corners in the DEM and the query image by exhaustively combining pairs of point-ray features. Then, we employ the two-point method to estimate both the camera pose and compute correspondences between buildings in the DEM and the query image. Finally, we show that the computed camera poses can be efficiently ranked by a simple skyline projection step using building edges from the DEM. Our experimental evaluation illustrates the promise of a purely geometric approach to the urban geo-localization problem.

Important diagnostic criteria for glaucoma are changes in the 3D structure of the optic disc due to optic nerve damage. We propose an automatic approach for detecting these changes in 3D models reconstructed from fundus images of the same patient taken at different times. For each time session, only two uncalibated fundus images are required. The approach applies a 6-point algorithm to estimate relative camera pose assuming a constant camera focal length. To deal with the instability of 3D reconstruction associated with fundus images, our approach keeps multiple candidate reconstruction solutions for each image pair. The best 3D reconstruction is found by optimizing the 3D registration of all images after an iterative bundle adjustment that tolerates possible structure changes. The 3D structure changes are detected by evaluating the reprojection errors of feature points in image space. We validate the approach by comparing the diagnosis results with manual grading by human experts on a fundus image dataset.

This paper presents a novel method to recover 3D structure of the optic disc in the retina from two uncalibrated fundus images. Retinal images are commonly uncalibrated when acquired clinically, creating rectification challenges as well as significant radiometric and blur differences within the stereo pair. By exploiting structural peculiarities of the retina, we modified the Graph Cuts computational stereo method (one of current state-of-the-art methods) to yield a high quality algorithm for fundus stereo reconstruction. Extensive qualitative and quantitative experimental evaluation (where OCT scans are used as 3D ground truth) on our and publicly available datasets shows the superiority of the proposed method in comparison to other alternatives.

Despite recent advances, automatic blood vessel extraction from low quality retina images remains difficult. We propose an interactive approach that enables a user to efficiently obtain near perfect vessel segmentation with a few mouse clicks. Given two seed points, the approach seeks an optimal path between them by minimizing a cost function. In contrast to the Live-Vessel approach, the graph in our approach is based on the curve fragments generated with vessel tracing instead of individual pixels. This enables our approach to overcome the shortcut problem in extracting tortuous vessels and the problem of vessel interference in extracting neighboring vessels in minimal-cost path techniques, resulting in less user interaction for extracting thin and tortuous vessels from low contrast images. It also makes the approach much faster.

We propose an adaptive rendering approach for large-scale skyline characterization and matching with applications to automated geo-tagging of photos and images. Given an image, our system automatically extracts the skyline and then matches it to a database of reference skylines extracted from rendered images using digital elevation data (DEM). The sampling density of these rendering locations determines both the accuracy and the speed of skyline matching. The proposed approach successfully combines global planning and local greedy search strategies to select new rendering locations incrementally. We report quantitative and qualitative results from synthesized and real experiments, where we achieve a computational speedup of around 4X.

We study the feasibility of solving the challenging problem of geolocalizing ground level images in urban areas with respect to a database of images captured from the air such as satellite and oblique aerial images. We observe that comprehensive aerial image databases are widely available while complete coverage of urban areas from the ground is at best spotty. As a result, localization of ground level imagery with respect to aerial collections is a technically important and practically significant problem. We exploit two key insights: (1) satellite image to oblique aerial image correspondences are used to extract building facades, and (2) building facades are matched between oblique aerial and ground images for geo-localization. Key contributions include: (1) A novel method for extracting building facades using building outlines; (2) Correspondence of building facades between oblique aerial and ground images without direct matching; and (3) Position and orientation estimation of ground images. We show results of ground image localization in a dense urban area.

Analyzing change in the 3D structure of the optic disc over time has long been recognized as central to the diagnosis of glaucoma but has been inadequately addressed by computer vision methods. Currently, clinicians examine stereo pairs from different time instants for interval changes indicative of glaucoma. Due to the clinical procedures in capturing optic disc images, these stereo pairs are usually completely uncalibrated - the camera intrinsics and extrinsics are unknown. Clinicians have to account for these unknown factors and hence their diagnoses of optic disc stability or change are subjective. Changes in the 3D structure of the optic disc are typically accompanied by changes in the 3D structure of blood vessels in that region. Therefore, change in the 3D structure of blood vessels can be used for glaucoma diagnosis. In this paper, we introduce a projective geometry based approach that reconstructs and aligns 3D blood vessel networks given two stereo pairs of optic disc images. We demonstrate that this alignment can identify regions where the vessel structure has changed. Since calibration is unavailable, the 3D structures and the alignment have a projective ambiguity, and hence, we cannot use an absolute threshold on the alignment error to automatically identify change. We have therefore developed an interactive tool that highlights regions with the largest alignment errors. This tool demonstrates the utility of our approach and also can guide clinical observers to optic disc regions where they should look for changes. We believe that our approach can serve as a platform to develop much needed novel tools for glaucoma diagnosis.

We present a novel LIDAR streaming architecture for real-time, on-board processing using unmanned robots. We propose a two-level 3D data structure that allows pipelined and streaming processing of the 3D data as it arrives from a moving robot: (i) at the coarse level, the incoming 3D scans are stored in memory in a dense 3D voxel grid with a relatively large voxel size - this ensures buffering of the most recent data and the availability of sufficient 3D measurements within a specific processing volume at the next level; (ii) at the fine level, we employ a data chunking mechanism guided by the movement of the robot and a rolling dense 3D voxel grid for processing the data in the immediate vicinity of the robot, which enables reuse of previously computed features. The architecture proposed requires a very small memory footprint, minimal data copying, and allows a fast spatial access for 3D data, even at the finest resolutions. We illustrate the proposed streaming architecture on a real-time 3D structure characterization task for detecting doors and stairs in indoor environments and show qualitative results demonstrating the effectiveness of our approach.

Localizing blood vessels in eye images is a crucial step in the automated and objective diagnosis of eye diseases. Most previous research has focused on extracting the centerlines of vessels in large field of view images. However, for diagnosing diseases of the optic disk region, like glaucoma, small field of view images have to be analyzed. One needs to identify not only the centerlines, but also vessel widths, which vary widely in these images. We present an automatic technique for localizing vessels in small field of view images using multi-scale matched filters. We also estimate local vessel properties - width and orientation - along the length of each vessel. Furthermore, we explicitly account for highlights on thick vessels - central reflexes - which are ignored in many previous works. Qualitative and quantitative results demonstrate the efficacy of our method - e.g. vessel centers are localized with RMS and median errors of 2.11 and 1 pixels, respectively in 700×700 images.

We present a real-time pedestrian detection system based on structure and appearance classification. We discuss several novel ideas that contribute to having low-false alarms and high detection rates, while at the same time achieving computational efficiency: (i) At the front end of our system we employ stereo to detect pedestrians in 3D range maps using template matching with a representative 3D shape model, and to classify other background objects in the scene such as buildings, trees and poles. The structure classification efficiently labels substantial amount of non-relevant image regions and guides the further computationally expensive process to focus on relatively small image parts; (ii)We improve the appearance-based classifiers based on HoG descriptors by performing template matching with 2D human shape contour fragments that results in improved localization and accuracy; (iii) We build a suite of classifiers tuned to specific distance ranges for optimized system performance. Our method is evaluated on publicly available datasets and is shown to match or exceed the performance of leading pedestrian detectors in terms of accuracy as well as achieving real-time computation (10 Hz), which makes it adequate for in-vehicle navigation platform.

We present an on-the-move LIDAR-based object detection system for autonomous and semi-autonomous unmanned vehicle systems. In this paper we make several contributions: (i) we describe an algorithm for real-time detection of objects such as doors and stairs in indoor environments; (ii) we describe efficient data structures and algorithms for processing 3D point clouds acquired by laser scanners in a streaming manner, which minimize the memory copying and access. We show qualitative results demonstrating the effectiveness of our approach on runs in an indoor office environment.

We present a real-time pedestrian detection system which uses cues derived from structure and appearance classification We discuss several novel ideas to achieve computational efficien y while improving on both detection and false-alarm rates: (i) At the front end of our system we employ stereo to detect pedestrians in 3D range maps, and to classify surrounding structure such as buildings, trees, poles etc. in the scene. The structure classificatio efficientl labels substantial amount of non-relevant image regions and guides the further computationally expensive process to focus on relatively small image parts; (ii) We improve the appearance-based classifier based on HoG descriptors by performing template matching with 2D human shape contour fragments that results in improved localization and accuracy; (iii) We train individual classifier at several depth ranges that allow us to account for appearance and 2D shape changes at variable distances in front of the camera. Our method is evaluated on publicly available datasets and is shown to match or exceed the performance of leading pedestrian detectors in terms of accuracy as well as achieving real-time computation (10 Hz), which makes it adequate for deployment in fiel robots and other navigation platforms.

We present a new multi-stage algorithm for car and truck detection from a moving vehicle. The algorithm performs a search for pertinent features in three dimensions, guided by a ground plane and lane boundary estimation sub-system, and assembles these features into vehicle hypotheses. A number of classifiers are applied to the hypotheses in order to remove false detections. Quantitative analysis on real-world test data show a detection rate of 99.4% and a false positive rate of 1.77%; a result that compares favourably with other systems in the literature.

This paper addresses the frame-to-frame data association and state estimation problems in localization of a pedestrian relative to a moving vehicle from a monocular far infra-red video sequence. Using a novel application of the hierarchical model-based motion estimation framework, we are able to use the image appearance information to solve the frame-to-frame data association problem and estimate a sub-pixel accurate height ratio for a pedestrian in two frames. Then, to localize the pedestrian, we propose a novel approach of using the pedestrian height ratio estimates to guide an interacting multiple-hypothesis-mode/height filtering algorithm instead of using a constant pedestrian height model. Experiments on several IR sequences demonstrate that this approach achieves results comparable to those from a known pedestrian height thus avoiding errors from a constant height model based approach.

Development of a practical stereo vision sensor for real-world applications must account for the variability of high-volume production processes and the impact of unknown environmental conditions during its operation. One critical factor of stereo depth estimation performance is the relative alignment of the stereo camera pair. While imperfectly aligned stereo cameras may be rectified in the image domain, there are some errors introduced by both the calibration recovery and image rectification processes. Finally, additional uncalibrated misalignments, for example due to thermal or mechanical deformation in a harsh automotive environment, may occur which will further deteriorate stereo depth estimation. This paper describes an experimental framework for determining these limits using image processing algorithms, operating on graphically synthesized imagery, with performance envelope validation on real stereo image data.

Genetic regulatory networks (GRNs) control gene expression and are responsible for establishing the regular cellular patterns that constitute an organism. This paper introduces a model of biological development that generates cellular patterns via chemical interactions. GRNs for protein expression are generated and evaluated for their effectiveness in constructing 2D patterns of cells such as borders, patches, and mosaics. Three types of searches were performed: (a) a Monte Carlo search of the GRN space using a utility function based on spatial interestingness; (b) a hill climbing search to identify GRNs that solve specific pattern problems; (c) a search for combinatorial codes that solve difficult target patterns by running multiple disjoint GRNs in parallel. We show that simple biologically realistic GRNs can construct many complex cellular patterns. Our model provides an avenue to explore the evolution of complex GRNs that drive development.

Glaucoma diagnosis, management, and research require complex assessments of the optic disc. Over many decades, the gold standard of optic disc evaluation in glaucoma has been the qualitative evaluation of stereo photographic images of the optic disc, most typically as 35-mm slides or other film-based media. More recently, automated optic disc analyzers have been introduced that use either laser technologies or polarized light to image the optic disc and/or the retinal nerve fiber layer. Studies with these instruments infrequently compare them directly to stereo photographs, and available reports generally find their diagnostic precision comparable to stereo photographs.1–5 So far, the qualitative evaluation of stereo optic disc images by experienced observers remains the reference standard for optic disc evaluation in glaucoma6–8 despite the subjectivity and variability inherent in this conventional approach.9,10 As examples, contemporary clinical glaucoma trials, such as the Ocular Hypertension Treatment Study (OHTS),11,12 the Low-Pressure Glaucoma Treatment Study (LoGTS)13 and the European Glaucoma Prevention Study (EGPS),14,15 used human observers to grade conventional stereo photos to evaluate the optic disc. We are adapting modern computer vision methods to optic disc diagnosis in glaucoma but were concerned with the limited number of published comparisons of digital and film imaging formats for optic disc photographs, as ophthalmic photography has migrated from film-based to digital images. The few published reports mostly have assessed the utility of digital optic nerve images for quantitative parameters long used in glaucoma evaluation, such as the cup-to-disc ratio.16–21 In contrast, in a considerably more extensive body of literature, the utility of digital images has been compared to that of film images in retinal diseases such as diabetic retinopathy22–26 or macular degeneration.27–29 To address the suitability of digital stereo optic disc images as a platform for developing novel computerized approaches for glaucoma diagnosis, we assessed the quality and definition of optic disc features by directly comparing optic disc images obtained from individual subjects and displayed in different formats. We sought to avoid the well-known inaccuracies in trying to assign quantitative estimates such as the cup-to-disc ratio.9,10 Specifically, we obtained stereo color optic disc images contemporaneously in both digital and 35-mm slide film formats in subjects with a range of optic disc cup sizes and included both scanned and grayscale images in the analysis.