Continual learning for computer vision applications

One of the most visionary goals of Artificial Intelligence is to create a system able to mimic and eventually surpass the intelligence observed in biological systems including, ambitiously, the one observed in humans. The main distinctive strength of humans is their ability to build a deep understanding of the world by learning continuously and drawing from their experiences. This ability, which is found in various degrees in all intelligent biological beings, allows them to adapt and properly react to changes by incrementally expanding and refining their knowledge. Arguably, achieving this ability is one of the main goals of Artificial Intelligence and a cornerstone towards the creation of intelligent artificial agents. Modern Deep Learning approaches allowed researchers and industries to achieve great advancements towards the resolution of many long-standing problems in areas like Computer Vision and Natural Language Processing. However, while this current age of renewed interest in AI allowed for the creation of extremely useful applications, a concerningly limited effort is being directed towards the design of systems able to learn continuously. The biggest problem that hinders an AI system from learning incrementally is the catastrophic forgetting phenomenon. This phenomenon, which was discovered in the 90s, naturally occurs in Deep Learning architectures where classic learning paradigms are applied when learning incrementally from a stream of experiences. This dissertation revolves around the Continual Learning field, a sub-field of Machine Learning research that has recently made a comeback following the renewed interest in Deep Learning approaches. This work will focus on a comprehensive view of continual learning by considering algorithmic, benchmarking, and applicative aspects of this field. This dissertation will also touch on community aspects such as the design and creation of research tools aimed at supporting Continual Learning research, and the theoretical and practical aspects concerning public competitions in this field.

Pattern recognition analysis on heavy ion reaction data

One of the problems in the analysis of nucleus-nucleus collisions is to get information on the value of the impact parameter b. This work consists in the application of pattern recognition techniques aimed at associating values of b to groups of events. To this end, a support vec- tor machine (SVM) classifier is adopted to analyze multifragmentation reactions. This method allows to backtracing the values of b through a particular multidimensional analysis. The SVM classification con- sists of two main phase. In the first one, known as training phase, the classifier learns to discriminate the events that are generated by two different model:Classical Molecular Dynamics (CMD) and Heavy- Ion Phase-Space Exploration (HIPSE) for the reaction: 58Ni +48 Ca at 25 AMeV. To check the classification of events in the second one, known as test phase, what has been learned is tested on new events generated by the same models. These new results have been com- pared to the ones obtained through others techniques of backtracing the impact parameter. Our tests show that, following this approach, the central collisions and peripheral collisions, for the CMD events, are always better classified with respect to the classification by the others techniques of backtracing. We have finally performed the SVM classification on the experimental data measured by NUCL-EX col- laboration with CHIMERA apparatus for the previous reaction.

The posterior parietal cortex: a bridge between vision and action

The present work takes into account three posterior parietal areas, V6, V6A, and PEc, all operating on different subsets of signals (visual, somatic, motor). The work focuses on the study of their functional properties, to better understand their respective contribution in the neuronal circuits that make possible the interactions between subject and external environment. In the caudalmost pole of parietal lobe there is area V6. Functional data suggest that this area is related to the encoding of both objects motion and ego-motion. However, the sensitivity of V6 neurons to optic flow stimulations has been tested only in human fMRI experiments. Here we addressed this issue by applying on monkey the same experimental protocol used in human studies. The visual stimulation obtained with the Flow Fields stimulus was the most effective and powerful to activate area V6 in monkey, further strengthening this homology between the two primates. The neighboring areas, V6A and PEc, show different cytoarchitecture and connectivity profiles, but are both involved in the control of reaches. We studied the sensory responses present in these areas, and directly compared these.. We also studied the motor related discharges of PEc neurons during reaching movements in 3D space comparing also the direction and depth tuning of PEc cells with those of V6A. The results show that area PEc and V6A share several functional properties. Area PEc, unlike V6A, contains a richer and more complex somatosensory input, and a poorer, although complex visual one. Differences emerged also comparing the motor-related properties for reaches in depth: the incidence of depth modulations in PEc and the temporal pattern of modulation for depth and direction allow to delineate a trend among the two parietal visuomotor areas.

Contribution of Vision and Proprioception to the Precision of Reaching Movements

Ren and colleagues (2006) found that saccades to visual targets became less accurate when somatosensory information about hand location was added, suggesting that saccades rely mainly on vision. We conducted two kinematic experiments to examine whether or not reaching movements would also show such strong reliance on vision. In Experiment 1, subjects used their dominant right hand to perform reaches, with or without a delay, to an external visual target or to their own left fingertip positioned either by the experimenter or by the participant. Unlike saccades, reaches became more accurate and precise when proprioceptive information was available. In Experiment 2, subjects reached toward external or bodily targets with differing amounts of visual information. Proprioception improved performance only when vision was limited. Our results indicate that reaching movements, unlike saccades, are improved rather than impaired by the addition of somatosensory information.

Engineering Agent-Oriented Technologies and Programming Languages for Computer Programming and Software Development

Mainstream hardware is becoming parallel, heterogeneous, and distributed on every desk, every home and in every pocket. As a consequence, in the last years software is having an epochal turn toward concurrency, distribution, interaction which is pushed by the evolution of hardware architectures and the growing of network availability. This calls for introducing further abstraction layers on top of those provided by classical mainstream programming paradigms, to tackle more effectively the new complexities that developers have to face in everyday programming. A convergence it is recognizable in the mainstream toward the adoption of the actor paradigm as a mean to unite object-oriented programming and concurrency. Nevertheless, we argue that the actor paradigm can only be considered a good starting point to provide a more comprehensive response to such a fundamental and radical change in software development. Accordingly, the main objective of this thesis is to propose Agent-Oriented Programming (AOP) as a high-level general purpose programming paradigm, natural evolution of actors and objects, introducing a further level of human-inspired concepts for programming software systems, meant to simplify the design and programming of concurrent, distributed, reactive/interactive programs. To this end, in the dissertation first we construct the required background by studying the state-of-the-art of both actor-oriented and agent-oriented programming, and then we focus on the engineering of integrated programming technologies for developing agent-based systems in their classical application domains: artificial intelligence and distributed artificial intelligence. Then, we shift the perspective moving from the development of intelligent software systems, toward general purpose software development. Using the expertise maturated during the phase of background construction, we introduce a general-purpose programming language named simpAL, which founds its roots on general principles and practices of software development, and at the same time provides an agent-oriented level of abstraction for the engineering of general purpose software systems.

Methodologies for visual correspondence

Visual correspondence is a key computer vision task that aims at identifying projections of the same 3D point into images taken either from different viewpoints or at different time instances. This task has been the subject of intense research activities in the last years in scenarios such as object recognition, motion detection, stereo vision, pattern matching, image registration. The approaches proposed in literature typically aim at improving the state of the art by increasing the reliability, the accuracy or the computational efficiency of visual correspondence algorithms. The research work carried out during the Ph.D. course and presented in this dissertation deals with three specific visual correspondence problems: fast pattern matching, stereo correspondence and robust image matching. The dissertation presents original contributions to the theory of visual correspondence, as well as applications dealing with 3D reconstruction and multi-view video surveillance.

Scene Mapping and Understanding by Robotic Vision

The first mechanical Automaton concept was found in a Chinese text written in the 3rd century BC, while Computer Vision was born in the late 1960s. Therefore, visual perception applied to machines (i.e. the Machine Vision) is a young and exciting alliance. When robots came in, the new field of Robotic Vision was born, and these terms began to be erroneously interchanged. In short, we can say that Machine Vision is an engineering domain, which concern the industrial use of Vision. The Robotic Vision, instead, is a research field that tries to incorporate robotics aspects in computer vision algorithms. Visual Servoing, for example, is one of the problems that cannot be solved by computer vision only. Accordingly, a large part of this work deals with boosting popular Computer Vision techniques by exploiting robotics: e.g. the use of kinematics to localize a vision sensor, mounted as the robot end-effector. The remainder of this work is dedicated to the counterparty, i.e. the use of computer vision to solve real robotic problems like grasping objects or navigate avoiding obstacles. Will be presented a brief survey about mapping data structures most widely used in robotics along with SkiMap, a novel sparse data structure created both for robotic mapping and as a general purpose 3D spatial index. Thus, several approaches to implement Object Detection and Manipulation, by exploiting the aforementioned mapping strategies, will be proposed, along with a completely new Machine Teaching facility in order to simply the training procedure of modern Deep Learning networks.