Segmentazione e categorizzazione di oggetti mediante immagini depth e deep learning

Questo lavoro è iniziato con uno studio teorico delle principali tecniche di classificazione di immagini note in letteratura, con particolare attenzione ai più diffusi modelli di rappresentazione dell’immagine, quali il modello Bag of Visual Words, e ai principali strumenti di Apprendimento Automatico (Machine Learning). In seguito si è focalizzata l’attenzione sulla analisi di ciò che costituisce lo stato dell’arte per la classificazione delle immagini, ovvero il Deep Learning. Per sperimentare i vantaggi dell’insieme di metodologie di Image Classification, si è fatto uso di Torch7, un framework di calcolo numerico, utilizzabile mediante il linguaggio di scripting Lua, open source, con ampio supporto alle metodologie allo stato dell’arte di Deep Learning. Tramite Torch7 è stata implementata la vera e propria classificazione di immagini poiché questo framework, grazie anche al lavoro di analisi portato avanti da alcuni miei colleghi in precedenza, è risultato essere molto efficace nel categorizzare oggetti in immagini. Le immagini su cui si sono basati i test sperimentali, appartengono a un dataset creato ad hoc per il sistema di visione 3D con la finalità di sperimentare il sistema per individui ipovedenti e non vedenti; in esso sono presenti alcuni tra i principali ostacoli che un ipovedente può incontrare nella propria quotidianità. In particolare il dataset si compone di potenziali ostacoli relativi a una ipotetica situazione di utilizzo all’aperto. Dopo avere stabilito dunque che Torch7 fosse il supporto da usare per la classificazione, l’attenzione si è concentrata sulla possibilità di sfruttare la Visione Stereo per aumentare l’accuratezza della classificazione stessa. Infatti, le immagini appartenenti al dataset sopra citato sono state acquisite mediante una Stereo Camera con elaborazione su FPGA sviluppata dal gruppo di ricerca presso il quale è stato svolto questo lavoro. Ciò ha permesso di utilizzare informazioni di tipo 3D, quali il livello di depth (profondità) di ogni oggetto appartenente all’immagine, per segmentare, attraverso un algoritmo realizzato in C++, gli oggetti di interesse, escludendo il resto della scena. L’ultima fase del lavoro è stata quella di testare Torch7 sul dataset di immagini, preventivamente segmentate attraverso l’algoritmo di segmentazione appena delineato, al fine di eseguire il riconoscimento della tipologia di ostacolo individuato dal sistema.

Deep learning based style transfer for low altitude aerial imagery

Unmanned Aerial Vehicle (UAVs) equipped with cameras have been fast deployed to a wide range of applications, such as smart cities, agriculture or search and rescue applications. Even though UAV datasets exist, the amount of open and quality UAV datasets is limited. So far, we want to overcome this lack of high quality annotation data by developing a simulation framework for a parametric generation of synthetic data. The framework accepts input via a serializable format. The input specifies which environment preset is used, the objects to be placed in the environment along with their position and orientation as well as additional information such as object color and size. The result is an environment that is able to produce UAV typical data: RGB image from the UAVs camera, altitude, roll, pitch and yawn of the UAV. Beyond the image generation process, we improve the resulting image data photorealism by using Synthetic-To-Real transfer learning methods. Transfer learning focuses on storing knowledge gained while solving one problem and applying it to a different - although related - problem. This approach has been widely researched in other affine fields and results demonstrate it to be an interesing area to investigate. Since simulated images are easy to create and synthetic-to-real translation has shown good quality results, we are able to generate pseudo-realistic images. Furthermore, object labels are inherently given, so we are capable of extending the already existing UAV datasets with realistic quality images and high resolution meta-data. During the development of this thesis we have been able to produce a result of 68.4% on UAVid. This can be considered a new state-of-art result on this dataset.

Depth Estimation in Stereo Biomedical Images via Proxy-Supervised Deep Learning

In order to estimate depth through supervised deep learning-based stereo methods, it is necessary to have access to precise ground truth depth data. While the gathering of precise labels is commonly tackled by deploying depth sensors, this is not always a viable solution. For instance, in many applications in the biomedical domain, the choice of sensors capable of sensing depth at small distances with high precision on difficult surfaces (that present non-Lambertian properties) is very limited. It is therefore necessary to find alternative techniques to gather ground truth data without having to rely on external sensors. In this thesis, two different approaches have been tested to produce supervision data for biomedical images. The first aims to obtain input stereo image pairs and disparities through simulation in a virtual environment, while the second relies on a non-learned disparity estimation algorithm in order to produce noisy disparities, which are then filtered by means of hand-crafted confidence measures to create noisy labels for a subset of pixels. Among the two, the second approach, which is referred in literature as proxy-labeling, has shown the best results and has even outperformed the non-learned disparity estimation algorithm used for supervision.

Deep learning models for the automatic analysis of volleyball plays

Collecting and analysing data is an important element in any field of human activity and research. Even in sports, collecting and analyzing statistical data is attracting a growing interest. Some exemplar use cases are: improvement of technical/tactical aspects for team coaches, definition of game strategies based on the opposite team play or evaluation of the performance of players. Other advantages are related to taking more precise and impartial judgment in referee decisions: a wrong decision can change the outcomes of important matches. Finally, it can be useful to provide better representations and graphic effects that make the game more engaging for the audience during the match. Nowadays it is possible to delegate this type of task to automatic software systems that can use cameras or even hardware sensors to collect images or data and process them. One of the most efficient methods to collect data is to process the video images of the sporting event through mixed techniques concerning machine learning applied to computer vision. As in other domains in which computer vision can be applied, the main tasks in sports are related to object detection, player tracking, and to the pose estimation of athletes. The goal of the present thesis is to apply different models of CNNs to analyze volleyball matches. Starting from video frames of a volleyball match, we reproduce a bird's eye view of the playing court where all the players are projected, reporting also for each player the type of action she/he is performing.

A deep learning-based approach for 3D people tracking

Questa tesi si occupa dell’estensione di un framework software finalizzato all'individuazione e al tracciamento di persone in una scena ripresa da telecamera stereoscopica. In primo luogo è rimossa la necessità di una calibrazione manuale offline del sistema sfruttando algoritmi che consentono di individuare, a partire da un fotogramma acquisito dalla camera, il piano su cui i soggetti tracciati si muovono. Inoltre, è introdotto un modulo software basato su deep learning con lo scopo di migliorare la precisione del tracciamento. Questo componente, che è in grado di individuare le teste presenti in un fotogramma, consente ridurre i dati analizzati al solo intorno della posizione effettiva di una persona, escludendo oggetti che l’algoritmo di tracciamento sarebbe portato a individuare come persone.

Deep learning of representations and its application to computer vision

L’objectif de cette thèse par articles est de présenter modestement quelques étapes du parcours qui mènera (on espère) à une solution générale du problème de l’intelligence artificielle. Cette thèse contient quatre articles qui présentent chacun une différente nouvelle méthode d’inférence perceptive en utilisant l’apprentissage machine et, plus particulièrement, les réseaux neuronaux profonds. Chacun de ces documents met en évidence l’utilité de sa méthode proposée dans le cadre d’une tâche de vision par ordinateur. Ces méthodes sont applicables dans un contexte plus général, et dans certains cas elles on tété appliquées ailleurs, mais ceci ne sera pas abordé dans le contexte de cette de thèse. Dans le premier article, nous présentons deux nouveaux algorithmes d’inférence variationelle pour le modèle génératif d’images appelé codage parcimonieux “spike- and-slab” (CPSS). Ces méthodes d’inférence plus rapides nous permettent d’utiliser des modèles CPSS de tailles beaucoup plus grandes qu’auparavant. Nous démontrons qu’elles sont meilleures pour extraire des détecteur de caractéristiques quand très peu d’exemples étiquetés sont disponibles pour l’entraînement. Partant d’un modèle CPSS, nous construisons ensuite une architecture profonde, la machine de Boltzmann profonde partiellement dirigée (MBP-PD). Ce modèle a été conçu de manière à simplifier d’entraînement des machines de Boltzmann profondes qui nécessitent normalement une phase de pré-entraînement glouton pour chaque couche. Ce problème est réglé dans une certaine mesure, mais le coût d’inférence dans le nouveau modèle est relativement trop élevé pour permettre de l’utiliser de manière pratique. Dans le deuxième article, nous revenons au problème d’entraînement joint de machines de Boltzmann profondes. Cette fois, au lieu de changer de famille de modèles, nous introduisons un nouveau critère d’entraînement qui donne naissance aux machines de Boltzmann profondes à multiples prédictions (MBP-MP). Les MBP-MP sont entraînables en une seule étape et ont un meilleur taux de succès en classification que les MBP classiques. Elles s’entraînent aussi avec des méthodes variationelles standard au lieu de nécessiter un classificateur discriminant pour obtenir un bon taux de succès en classification. Par contre, un des inconvénients de tels modèles est leur incapacité de générer deséchantillons, mais ceci n’est pas trop grave puisque la performance de classification des machines de Boltzmann profondes n’est plus une priorité étant donné les dernières avancées en apprentissage supervisé. Malgré cela, les MBP-MP demeurent intéressantes parce qu’elles sont capable d’accomplir certaines tâches que des modèles purement supervisés ne peuvent pas faire, telles que celle de classifier des données incomplètes ou encore celle de combler intelligemment l’information manquante dans ces données incomplètes. Le travail présenté dans cette thèse s’est déroulé au milieu d’une période de transformations importantes du domaine de l’apprentissage à réseaux neuronaux profonds qui a été déclenchée par la découverte de l’algorithme de “dropout” par Geoffrey Hinton. Dropout rend possible un entraînement purement supervisé d’architectures de propagation unidirectionnel sans être exposé au danger de sur- entraînement. Le troisième article présenté dans cette thèse introduit une nouvelle fonction d’activation spécialement con ̧cue pour aller avec l’algorithme de Dropout. Cette fonction d’activation, appelée maxout, permet l’utilisation de aggrégation multi-canal dans un contexte d’apprentissage purement supervisé. Nous démontrons comment plusieurs tâches de reconnaissance d’objets sont mieux accomplies par l’utilisation de maxout. Pour terminer, sont présentons un vrai cas d’utilisation dans l’industrie pour la transcription d’adresses de maisons à plusieurs chiffres. En combinant maxout avec une nouvelle sorte de couche de sortie pour des réseaux neuronaux de convolution, nous démontrons qu’il est possible d’atteindre un taux de succès comparable à celui des humains sur un ensemble de données coriace constitué de photos prises par les voitures de Google. Ce système a été déployé avec succès chez Google pour lire environ cent million d’adresses de maisons.

Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.

Computer Vision-Based Carbohydrate Estimation for Type 1 Diabetic Patients Using Smartphones

Background: Individuals with type 1 diabetes (T1D) have to count the carbohydrates (CHOs) of their meal to estimate the prandial insulin dose needed to compensate for the meal’s effect on blood glucose levels. CHO counting is very challenging but also crucial, since an error of 20 grams can substantially impair postprandial control. Method: The GoCARB system is a smartphone application designed to support T1D patients with CHO counting of nonpacked foods. In a typical scenario, the user places a reference card next to the dish and acquires 2 images with his/her smartphone. From these images, the plate is detected and the different food items on the plate are automatically segmented and recognized, while their 3D shape is reconstructed. Finally, the food volumes are calculated and the CHO content is estimated by combining the previous results and using the USDA nutritional database. Results: To evaluate the proposed system, a set of 24 multi-food dishes was used. For each dish, 3 pairs of images were taken and for each pair, the system was applied 4 times. The mean absolute percentage error in CHO estimation was 10 ± 12%, which led to a mean absolute error of 6 ± 8 CHO grams for normal-sized dishes. Conclusion: The laboratory experiments demonstrated the feasibility of the GoCARB prototype system since the error was below the initial goal of 20 grams. However, further improvements and evaluation are needed prior launching a system able to meet the inter- and intracultural eating habits.

Learning and Exploiting Camera Geometry for Computer Vision

This work explores the use of statistical methods in describing and estimating camera poses, as well as the information feedback loop between camera pose and object detection. Surging development in robotics and computer vision has pushed the need for algorithms that infer, understand, and utilize information about the position and orientation of the sensor platforms when observing and/or interacting with their environment.

The first contribution of this thesis is the development of a set of statistical tools for representing and estimating the uncertainty in object poses. A distribution for representing the joint uncertainty over multiple object positions and orientations is described, called the mirrored normal-Bingham distribution. This distribution generalizes both the normal distribution in Euclidean space, and the Bingham distribution on the unit hypersphere. It is shown to inherit many of the convenient properties of these special cases: it is the maximum-entropy distribution with fixed second moment, and there is a generalized Laplace approximation whose result is the mirrored normal-Bingham distribution. This distribution and approximation method are demonstrated by deriving the analytical approximation to the wrapped-normal distribution. Further, it is shown how these tools can be used to represent the uncertainty in the result of a bundle adjustment problem.

Another application of these methods is illustrated as part of a novel camera pose estimation algorithm based on object detections. The autocalibration task is formulated as a bundle adjustment problem using prior distributions over the 3D points to enforce the objects' structure and their relationship with the scene geometry. This framework is very flexible and enables the use of off-the-shelf computational tools to solve specialized autocalibration problems. Its performance is evaluated using a pedestrian detector to provide head and foot location observations, and it proves much faster and potentially more accurate than existing methods.

Finally, the information feedback loop between object detection and camera pose estimation is closed by utilizing camera pose information to improve object detection in scenarios with significant perspective warping. Methods are presented that allow the inverse perspective mapping traditionally applied to images to be applied instead to features computed from those images. For the special case of HOG-like features, which are used by many modern object detection systems, these methods are shown to provide substantial performance benefits over unadapted detectors while achieving real-time frame rates, orders of magnitude faster than comparable image warping methods.

The statistical tools and algorithms presented here are especially promising for mobile cameras, providing the ability to autocalibrate and adapt to the camera pose in real time. In addition, these methods have wide-ranging potential applications in diverse areas of computer vision, robotics, and imaging.

Computer vision in aseptic pharmaceutical manufacturing: a deep learning-based approach to vial inspection and pack formation during a lyophilizer loading cycle

Vision systems are powerful tools playing an increasingly important role in modern industry, to detect errors and maintain product standards. With the enlarged availability of affordable industrial cameras, computer vision algorithms have been increasingly applied in industrial manufacturing processes monitoring. Until a few years ago, industrial computer vision applications relied only on ad-hoc algorithms designed for the specific object and acquisition setup being monitored, with a strong focus on co-designing the acquisition and processing pipeline. Deep learning has overcome these limits providing greater flexibility and faster re-configuration. In this work, the process to be inspected consists in vials’ pack formation entering a freeze-dryer, which is a common scenario in pharmaceutical active ingredient packaging lines. To ensure that the machine produces proper packs, a vision system is installed at the entrance of the freeze-dryer to detect eventual anomalies with execution times compatible with the production specifications. Other constraints come from sterility and safety standards required in pharmaceutical manufacturing. This work presents an overview about the production line, with particular focus on the vision system designed, and about all trials conducted to obtain the final performance. Transfer learning, alleviating the requirement for a large number of training data, combined with data augmentation methods, consisting in the generation of synthetic images, were used to effectively increase the performances while reducing the cost of data acquisition and annotation. The proposed vision algorithm is composed by two main subtasks, designed respectively to vials counting and discrepancy detection. The first one was trained on more than 23k vials (about 300 images) and tested on 5k more (about 75 images), whereas 60 training images and 52 testing images were used for the second one.

Optimized aperture shapes for depth estimation

The finite depth of field of a real camera can be used to estimate the depth structure of a scene. The distance of an object from the plane in focus determines the defocus blur size. The shape of the blur depends on the shape of the aperture. The blur shape can be designed by masking the main lens aperture. In fact, aperture shapes different from the standard circular aperture give improved accuracy of depth estimation from defocus blur. We introduce an intuitive criterion to design aperture patterns for depth from defocus. The criterion is independent of a specific depth estimation algorithm. We formulate our design criterion by imposing constraints directly in the data domain and optimize the amount of depth information carried by blurred images. Our criterion is a quadratic function of the aperture transmission values. As such, it can be numerically evaluated to estimate optimized aperture patterns quickly. The proposed mask optimization procedure is applicable to different depth estimation scenarios. We use it for depth estimation from two images with different focus settings, for depth estimation from two images with different aperture shapes as well as for depth estimation from a single coded aperture image. In this work we show masks obtained with this new evaluation criterion and test their depth discrimination capability using a state-of-the-art depth estimation algorithm.

Parallel Computer Vision Algorithms for Graphics Processing Units

La evolución de los teléfonos móviles inteligentes, dotados de cámaras digitales, está provocando una creciente demanda de aplicaciones cada vez más complejas que necesitan algoritmos de visión artificial en tiempo real; puesto que el tamaño de las señales de vídeo no hace sino aumentar y en cambio el rendimiento de los procesadores de un solo núcleo se ha estancado, los nuevos algoritmos que se diseñen para visión artificial han de ser paralelos para poder ejecutarse en múltiples procesadores y ser computacionalmente escalables. Una de las clases de procesadores más interesantes en la actualidad se encuentra en las tarjetas gráficas (GPU), que son dispositivos que ofrecen un alto grado de paralelismo, un excelente rendimiento numérico y una creciente versatilidad, lo que los hace interesantes para llevar a cabo computación científica. En esta tesis se exploran dos aplicaciones de visión artificial que revisten una gran complejidad computacional y no pueden ser ejecutadas en tiempo real empleando procesadores tradicionales. En cambio, como se demuestra en esta tesis, la paralelización de las distintas subtareas y su implementación sobre una GPU arrojan los resultados deseados de ejecución con tasas de refresco interactivas. Asimismo, se propone una técnica para la evaluación rápida de funciones de complejidad arbitraria especialmente indicada para su uso en una GPU. En primer lugar se estudia la aplicación de técnicas de síntesis de imágenes virtuales a partir de únicamente dos cámaras lejanas y no paralelas—en contraste con la configuración habitual en TV 3D de cámaras cercanas y paralelas—con información de color y profundidad. Empleando filtros de mediana modificados para la elaboración de un mapa de profundidad virtual y proyecciones inversas, se comprueba que estas técnicas son adecuadas para una libre elección del punto de vista. Además, se demuestra que la codificación de la información de profundidad con respecto a un sistema de referencia global es sumamente perjudicial y debería ser evitada. Por otro lado se propone un sistema de detección de objetos móviles basado en técnicas de estimación de densidad con funciones locales. Este tipo de técnicas es muy adecuada para el modelado de escenas complejas con fondos multimodales, pero ha recibido poco uso debido a su gran complejidad computacional. El sistema propuesto, implementado en tiempo real sobre una GPU, incluye propuestas para la estimación dinámica de los anchos de banda de las funciones locales, actualización selectiva del modelo de fondo, actualización de la posición de las muestras de referencia del modelo de primer plano empleando un filtro de partículas multirregión y selección automática de regiones de interés para reducir el coste computacional. Los resultados, evaluados sobre diversas bases de datos y comparados con otros algoritmos del estado del arte, demuestran la gran versatilidad y calidad de la propuesta. Finalmente se propone un método para la aproximación de funciones arbitrarias empleando funciones continuas lineales a tramos, especialmente indicada para su implementación en una GPU mediante el uso de las unidades de filtraje de texturas, normalmente no utilizadas para cómputo numérico. La propuesta incluye un riguroso análisis matemático del error cometido en la aproximación en función del número de muestras empleadas, así como un método para la obtención de una partición cuasióptima del dominio de la función para minimizar el error. ABSTRACT The evolution of smartphones, all equipped with digital cameras, is driving a growing demand for ever more complex applications that need to rely on real-time computer vision algorithms. However, video signals are only increasing in size, whereas the performance of single-core processors has somewhat stagnated in the past few years. Consequently, new computer vision algorithms will need to be parallel to run on multiple processors and be computationally scalable. One of the most promising classes of processors nowadays can be found in graphics processing units (GPU). These are devices offering a high parallelism degree, excellent numerical performance and increasing versatility, which makes them interesting to run scientific computations. In this thesis, we explore two computer vision applications with a high computational complexity that precludes them from running in real time on traditional uniprocessors. However, we show that by parallelizing subtasks and implementing them on a GPU, both applications attain their goals of running at interactive frame rates. In addition, we propose a technique for fast evaluation of arbitrarily complex functions, specially designed for GPU implementation. First, we explore the application of depth-image–based rendering techniques to the unusual configuration of two convergent, wide baseline cameras, in contrast to the usual configuration used in 3D TV, which are narrow baseline, parallel cameras. By using a backward mapping approach with a depth inpainting scheme based on median filters, we show that these techniques are adequate for free viewpoint video applications. In addition, we show that referring depth information to a global reference system is ill-advised and should be avoided. Then, we propose a background subtraction system based on kernel density estimation techniques. These techniques are very adequate for modelling complex scenes featuring multimodal backgrounds, but have not been so popular due to their huge computational and memory complexity. The proposed system, implemented in real time on a GPU, features novel proposals for dynamic kernel bandwidth estimation for the background model, selective update of the background model, update of the position of reference samples of the foreground model using a multi-region particle filter, and automatic selection of regions of interest to reduce computational cost. The results, evaluated on several databases and compared to other state-of-the-art algorithms, demonstrate the high quality and versatility of our proposal. Finally, we propose a general method for the approximation of arbitrarily complex functions using continuous piecewise linear functions, specially formulated for GPU implementation by leveraging their texture filtering units, normally unused for numerical computation. Our proposal features a rigorous mathematical analysis of the approximation error in function of the number of samples, as well as a method to obtain a suboptimal partition of the domain of the function to minimize approximation error.

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.

Hand gesture recognition system based in computer vision and machine learning : applications on human-machine interaction

Sendo uma forma natural de interação homem-máquina, o reconhecimento de gestos implica uma forte componente de investigação em áreas como a visão por computador e a aprendizagem computacional. O reconhecimento gestual é uma área com aplicações muito diversas, fornecendo aos utilizadores uma forma mais natural e mais simples de comunicar com sistemas baseados em computador, sem a necessidade de utilização de dispositivos extras. Assim, o objectivo principal da investigação na área de reconhecimento de gestos aplicada à interacção homemmáquina é o da criação de sistemas, que possam identificar gestos específicos e usálos para transmitir informações ou para controlar dispositivos. Para isso as interfaces baseados em visão para o reconhecimento de gestos, necessitam de detectar a mão de forma rápida e robusta e de serem capazes de efetuar o reconhecimento de gestos em tempo real. Hoje em dia, os sistemas de reconhecimento de gestos baseados em visão são capazes de trabalhar com soluções específicas, construídos para resolver um determinado problema e configurados para trabalhar de uma forma particular. Este projeto de investigação estudou e implementou soluções, suficientemente genéricas, com o recurso a algoritmos de aprendizagem computacional, permitindo a sua aplicação num conjunto alargado de sistemas de interface homem-máquina, para reconhecimento de gestos em tempo real. A solução proposta, Gesture Learning Module Architecture (GeLMA), permite de forma simples definir um conjunto de comandos que pode ser baseado em gestos estáticos e dinâmicos e que pode ser facilmente integrado e configurado para ser utilizado numa série de aplicações. É um sistema de baixo custo e fácil de treinar e usar, e uma vez que é construído unicamente com bibliotecas de código. As experiências realizadas permitiram mostrar que o sistema atingiu uma precisão de 99,2% em termos de reconhecimento de gestos estáticos e uma precisão média de 93,7% em termos de reconhecimento de gestos dinâmicos. Para validar a solução proposta, foram implementados dois sistemas completos. O primeiro é um sistema em tempo real capaz de ajudar um árbitro a arbitrar um jogo de futebol robótico. A solução proposta combina um sistema de reconhecimento de gestos baseada em visão com a definição de uma linguagem formal, o CommLang Referee, à qual demos a designação de Referee Command Language Interface System (ReCLIS). O sistema identifica os comandos baseados num conjunto de gestos estáticos e dinâmicos executados pelo árbitro, sendo este posteriormente enviado para um interface de computador que transmite a respectiva informação para os robôs. O segundo é um sistema em tempo real capaz de interpretar um subconjunto da Linguagem Gestual Portuguesa. As experiências demonstraram que o sistema foi capaz de reconhecer as vogais em tempo real de forma fiável. Embora a solução implementada apenas tenha sido treinada para reconhecer as cinco vogais, o sistema é facilmente extensível para reconhecer o resto do alfabeto. As experiências também permitiram mostrar que a base dos sistemas de interação baseados em visão pode ser a mesma para todas as aplicações e, deste modo facilitar a sua implementação. A solução proposta tem ainda a vantagem de ser suficientemente genérica e uma base sólida para o desenvolvimento de sistemas baseados em reconhecimento gestual que podem ser facilmente integrados com qualquer aplicação de interface homem-máquina. A linguagem formal de definição da interface pode ser redefinida e o sistema pode ser facilmente configurado e treinado com um conjunto de gestos diferentes de forma a serem integrados na solução final.