ImAge: an extensible agent-based architecture for image retrieval

We present an open and extensible architecture, ImAge, for content-based image retrieval in a distributed environment. The architecture proposes the use of system components with standard public interfaces for implementing retrieval functionality. The standardization of the components and their encapsulation in autonomous software agents result in functional stratification and easy extensibility. Collaboration of the independent retrieval resources in ImAge results in enhanced system capability. Reuse of existing retrieval resources is achieved by encapsulating them in agents with standard interfaces. The addition of independent agents with domain knowledege adds the capability of processing conceptual queries, while reusing the existing system components for feature-based retrieval. A communication protocol allows the declaration of the capabilities of the system components and negotiations for optimal resource selection for solving a retrieval problem. The use of mobile agents alleviates network bottlenecks. This paper describes a prototype implementation that validates the architecture.     

Visually guided ranging from observations of points, lines and curves via an identifier based nonlinear observer

In this paper we consider the problem of estimating the range information of features on an affine plane in by observing its image with the aid of a CCD camera, wherein we assume that the camera is undergoing a known motion. The features considered are points, lines and planar curves located on planar surfaces of static objects. The dynamics of the moving projections of the features on the image plane have been described as a suitable differential equation on an appropriate feature space. This dynamics is used to estimate feature parameters from which the range information is readily available. In this paper the proposed identification has been carried out via a newly introduced identifier based observer. Performance of the observer has been studied via simulation.     

STDNet: A CNN-based approach to single-/mixed-script detection

Innovations in Systems and Software Engineering - Script identification serves as a guide to the detection of the text of the scene through optical character recognition (OCR). But this is not a...     

Deep Learning Guided Double Hidden Layer Neural Synchronization Through Mutual Learning

Neural Processing Letters - In this paper, a Double Hidden Layer Neural Networks synchronization mechanism using Generative Adversarial Network (GAN) and mutual learning is used for the development...     

Cell tracking in microscopic video using matching and linking of bipartite graphs

Automated visual tracking of cells from video microscopy has many important biomedical applications. In this paper, we track human monocyte cells in a fluorescent microscopic video using matching and linking of bipartite graphs. Tracking of cells over a pair of frames is modeled as a maximum cardinality minimum weight matching problem for a bipartite graph with a novel cost function. The tracking results are further refined using a rank-based filtering mechanism. Linking of cell trajectories over different frames are achieved through composition of bipartite matches. The proposed solution does not require any explicit motion model, is highly scalable, and, can effectively handle the entry and exit of cells. Our tracking accuracy of (97.97 ± 0.94)% is superior than several existing methods [(95.66 ± 2.39)% [11], (94.42 ± 2.08)% [20], (81.22 ± 5.75)% [13], (78.31 ± 4.70)% [14]] and is highly comparable (98.20 ± 1.22)% to a recently published algorithm [26].     

On hybrid classification using model assisted posterior estimates

Traditional parametric and nonparametric classifiers used for statistical pattern recognition have their own strengths and limitations. While parametric methods assume some specific parametric models for density functions or posterior probabilities of competing classes, nonparametric methods are free from such assumptions. So, when these model assumptions are correct, parametric methods outperform nonparametric classifiers, especially when the training sample is small. But, violations of these assumptions often lead to poor performance by parametric classifiers, where nonparametric methods work well. In this article, we make an attempt to overcome these limitations of parametric and nonparametric approaches and combine their strengths. The resulting classifiers, denoted the hybrid classifiers, perform like parametric classifiers when the model assumptions are valid, but unlike parametric classifiers, they also provide safeguards against possible deviations from parametric model assumptions. In this article, we propose some multiscale methods for hybrid classification, and their performance is evaluated using several simulated and benchmark data sets.     

The Design and Role of Synthetic Creative Traits in Artificial Ant Colonies

Previous and current research on ant colonies have focused on the emergence of complex and sophisticated group-behaviors that are characteristic of the whole colony, starting from simple rules characterizing each individual ant. Examples of emergent behaviors include foraging and nest building. While social scientists, biologists, and physicists have focused on observing and analyzing the behavior of actual ants, researchers in computer science have investigated through modeling and simulation of synthetic ants on the computer, the emergent behavior in a collection of simple robots, simulated evolution of computer programs, and the formation of pheromone patterns during foraging under unlimited food supply. This paper has two objectives. First, it presents a systematic study, through behavior modeling and simulation, of the influence of the presence or absence of pheromone, the duration interval of pheromone, the extent of the food supply at the food sources, the size of the colony, and the search strategy employed during foraging, for a given geometry and a finite number of food sources, on the performance of the ant colony. The second and most important objective of this paper is to scientifically study the nature of creativity by modeling synthetic creativity in an ant colony, simulating it on a computer system, and measuring its impact on performance through innovative metric design. The study is motivated by leading thinkers, throughout time, who have strongly emphasized the importance of creativity over intelligence. While an exact and universal definition of creativity is elusive, the important characteristics include originality, intuition, and imagination. Creativity is best understood through its manifestations as novel and radically different ideas that transcend current knowledge and reasoning. In this study, two synthetic creativity traits are introduced into select individual ants of a colony by imparting to them a foraging behavior that is radically different from the normal behavior. Under the first trait, the creative ants choose to ignore the existing pheromone trails and search for food sources. Under the second trait, a creative ant shares its knowledge of food source location, after discovery, with all other creative ants, so that together they can focus solely on discovering new food sources. Analysis of the simulation results reveal that a creative trait coupled with the underlying parameters of the ant colony may cause the foraging completion time metric, i.e. the time to collect food from all sources, to be marginally better or weaker than the normal colony. Contrary to intuition, the completion time metric worsens when, in a fixed sized colony, the number of creative ants, relative to normal ants, becomes excessive. Furthermore, for a given finite geometry, finite food supply, and a fixed ratio of the number of creative to normal ants, the impact of creativity on the foraging performance is dependent on the underlying parameters, and may either continue to improve, remain unchanged, or decrease, as the size of the colony increases. Finally, while the creative ants expend, on an average, two to three times more energy than the normal ants, the ant colony with the creative ants discovers more food sources, in less time and, under certain circumstances, a creative colony discovers all food sources which a normal colony can never achieve.     

A UML-based pattern specification technique

Informally described design patterns are useful for communicating proven solutions for recurring design problems to developers, but they cannot be used as compliance points against which solutions that claim to conform to the patterns are checked. Pattern specification languages that utilize mathematical notation provide the needed formality, but often at the expense of usability. We present a rigorous and practical technique for specifying pattern solutions expressed in the unified modeling language (UML). The specification technique paves the way for the development of tools that support rigorous application of design patterns to UML design models. The technique has been used to create specifications of solutions for several popular design patterns. We illustrate the use of the technique by specifying observer and visitor pattern solutions.