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.

Non invasive classification system of scoliosis curve types using least-squares support vector machines

Objective To determine scoliosis curve types using non invasive surface acquisition, without prior knowledge from X-ray data. Methods Classification of scoliosis deformities according to curve type is used in the clinical management of scoliotic patients. In this work, we propose a robust system that can determine the scoliosis curve type from non invasive acquisition of the 3D back surface of the patients. The 3D image of the surface of the trunk is divided into patches and local geometric descriptors characterizing the back surface are computed from each patch and constitute the features. We reduce the dimensionality by using principal component analysis and retain 53 components using an overlap criterion combined with the total variance in the observed variables. In this work, a multi-class classifier is built with least-squares support vector machines (LS-SVM). The original LS-SVM formulation was modified by weighting the positive and negative samples differently and a new kernel was designed in order to achieve a robust classifier. The proposed system is validated using data from 165 patients with different scoliosis curve types. The results of our non invasive classification were compared with those obtained by an expert using X-ray images. Results The average rate of successful classification was computed using a leave-one-out cross-validation procedure. The overall accuracy of the system was 95%. As for the correct classification rates per class, we obtained 96%, 84% and 97% for the thoracic, double major and lumbar/thoracolumbar curve types, respectively. Conclusion This study shows that it is possible to find a relationship between the internal deformity and the back surface deformity in scoliosis with machine learning methods. The proposed system uses non invasive surface acquisition, which is safe for the patient as it involves no radiation. Also, the design of a specific kernel improved classification performance.

Design and Development of an Adaptable Frame -Based System for Dravidian Language Processing

This work is aimed at building an adaptable frame-based system for processing Dravidian languages. There are about 17 languages in this family and they are spoken by the people of South India.Karaka relations are one of the most important features of Indian languages. They are the semabtuco-syntactic relations between verbs and other related constituents in a sentence. The karaka relations and surface case endings are analyzed for meaning extraction. This approach is comparable with the borad class of case based grammars.The efficiency of this approach is put into test in two applications. One is machine translation and the other is a natural language interface (NLI) for information retrieval from databases. The system mainly consists of a morphological analyzer, local word grouper, a parser for the source language and a sentence generator for the target language. This work make contributios like, it gives an elegant account of the relation between vibhakthi and karaka roles in Dravidian languages. This mapping is elegant and compact. The same basic thing also explains simple and complex sentence in these languages. This suggests that the solution is not just ad hoc but has a deeper underlying unity. This methodology could be extended to other free word order languages. Since the frame designed for meaning representation is general, they are adaptable to other languages coming in this group and to other applications.

Studies Using Computer Simulation Models to Solve Problems Related to Logistics Terminals

This thesis deals with the use of simulation as a problem-solving tool to solve a few logistic system related problems. More specifically it relates to studies on transport terminals. Transport terminals are key elements in the supply chains of industrial systems. One of the problems related to use of simulation is that of the multiplicity of models needed to study different problems. There is a need for development of methodologies related to conceptual modelling which will help reduce the number of models needed. Three different logistic terminal systems Viz. a railway yard, container terminal of apart and airport terminal were selected as cases for this study. The standard methodology for simulation development consisting of system study and data collection, conceptual model design, detailed model design and development, model verification and validation, experimentation, and analysis of results, reporting of finding were carried out. We found that models could be classified into tightly pre-scheduled, moderately pre-scheduled and unscheduled systems. Three types simulation models( called TYPE 1, TYPE 2 and TYPE 3) of various terminal operations were developed in the simulation package Extend. All models were of the type discrete-event simulation. Simulation models were successfully used to help solve strategic, tactical and operational problems related to three important logistic terminals as set in our objectives. From the point of contribution to conceptual modelling we have demonstrated that clubbing problems into operational, tactical and strategic and matching them with tightly pre-scheduled, moderately pre-scheduled and unscheduled systems is a good workable approach which reduces the number of models needed to study different terminal related problems.

Management of Traffic Control — A Computer Oriented Approach

Traffic Management system (TMS) comprises four major sub systems: The Network Database Management system for information to the passengers, Transit Facility Management System for service, planning, and scheduling vehicle and crews, Congestion Management System for traffic forecasting and planning, Safety Management System concerned with safety aspects of passengers and Environment. This work has opened a rather wide frame work of model structures for application on traffic. The facets of these theories are so wide that it seems impossible to present all necessary models in this work. However it could be deduced from the study that the best Traffic Management System is that whichis realistic in all aspects is easy to understand is easy to apply As it is practically difficult to device an ideal fool—proof model, the attempt here has been to make some progress-in that direction.

A Writer Identification System For Handwritten Malayalam Documents

Handwriting is an acquired tool used for communication of one's observations or feelings. Factors that inuence a person's handwriting not only dependent on the individual's bio-mechanical constraints, handwriting education received, writing instrument, type of paper, background, but also factors like stress, motivation and the purpose of the handwriting. Despite the high variation in a person's handwriting, recent results from different writer identification studies have shown that it possesses sufficient individual traits to be used as an identification method. Handwriting as a behavioral biometric has had the interest of researchers for a long time. But recently it has been enjoying new interest due to an increased need and effort to deal with problems ranging from white-collar crime to terrorist threats. The identification of the writer based on a piece of handwriting is a challenging task for pattern recognition. The main objective of this thesis is to develop a text independent writer identification system for Malayalam Handwriting. The study also extends to developing a framework for online character recognition of Grantha script and Malayalam characters

Design And Development Of A Named Entity Based Question Answering System For Malayalam Language

This is a Named Entity Based Question Answering System for Malayalam Language. Although a vast amount of information is available today in digital form, no effective information access mechanism exists to provide humans with convenient information access. Information Retrieval and Question Answering systems are the two mechanisms available now for information access. Information systems typically return a long list of documents in response to a user’s query which are to be skimmed by the user to determine whether they contain an answer. But a Question Answering System allows the user to state his/her information need as a natural language question and receives most appropriate answer in a word or a sentence or a paragraph. This system is based on Named Entity Tagging and Question Classification. Document tagging extracts useful information from the documents which will be used in finding the answer to the question. Question Classification extracts useful information from the question to determine the type of the question and the way in which the question is to be answered. Various Machine Learning methods are used to tag the documents. Rule-Based Approach is used for Question Classification. Malayalam belongs to the Dravidian family of languages and is one of the four major languages of this family. It is one of the 22 Scheduled Languages of India with official language status in the state of Kerala. It is spoken by 40 million people. Malayalam is a morphologically rich agglutinative language and relatively of free word order. Also Malayalam has a productive morphology that allows the creation of complex words which are often highly ambiguous. Document tagging tools such as Parts-of-Speech Tagger, Phrase Chunker, Named Entity Tagger, and Compound Word Splitter are developed as a part of this research work. No such tools were available for Malayalam language. Finite State Transducer, High Order Conditional Random Field, Artificial Immunity System Principles, and Support Vector Machines are the techniques used for the design of these document preprocessing tools. This research work describes how the Named Entity is used to represent the documents. Single sentence questions are used to test the system. Overall Precision and Recall obtained are 88.5% and 85.9% respectively. This work can be extended in several directions. The coverage of non-factoid questions can be increased and also it can be extended to include open domain applications. Reference Resolution and Word Sense Disambiguation techniques are suggested as the future enhancements

A Prototype for a Multimodal Biometric Security system based on Face and Audio Signatures

Any automatically measurable, robust and distinctive physical characteristic or personal trait that can be used to identify an individual or verify the claimed identity of an individual, referred to as biometrics, has gained significant interest in the wake of heightened concerns about security and rapid advancements in networking, communication and mobility. Multimodal biometrics is expected to be ultra-secure and reliable, due to the presence of multiple and independent—verification clues. In this study, a multimodal biometric system utilising audio and facial signatures has been implemented and error analysis has been carried out. A total of one thousand face images and 250 sound tracks of 50 users are used for training the proposed system. To account for the attempts of the unregistered signatures data of 25 new users are tested. The short term spectral features were extracted from the sound data and Vector Quantization was done using K-means algorithm. Face images are identified based on Eigen face approach using Principal Component Analysis. The success rate of multimodal system using speech and face is higher when compared to individual unimodal recognition systems

IRIS BIOMETRIC RECOGNITION SYSTEM EMPLOYING CANNY OPERATOR

Biometrics has become important in security applications. In comparison with many other biometric features, iris recognition has very high recognition accuracy because it depends on iris which is located in a place that still stable throughout human life and the probability to find two identical iris's is close to zero. The identification system consists of several stages including segmentation stage which is the most serious and critical one. The current segmentation methods still have limitation in localizing the iris due to circular shape consideration of the pupil. In this research, Daugman method is done to investigate the segmentation techniques. Eyelid detection is another step that has been included in this study as a part of segmentation stage to localize the iris accurately and remove unwanted area that might be included. The obtained iris region is encoded using haar wavelets to construct the iris code, which contains the most discriminating feature in the iris pattern. Hamming distance is used for comparison of iris templates in the recognition stage. The dataset which is used for the study is UBIRIS database. A comparative study of different edge detector operator is performed. It is observed that canny operator is best suited to extract most of the edges to generate the iris code for comparison. Recognition rate of 89% and rejection rate of 95% is achieved

A model of Distributed Ayurvedic Diagnosis and Therapy System for Hridroga Using Agents

Diagnosis of Hridroga (cardiac disorders) in Ayurveda requires the combination of many different types of data, including personal details, patient symptoms, patient histories, general examination results, Ashtavidha pareeksha results etc. Computer-assisted decision support systems must be able to combine these data types into a seamless system. Intelligent agents, an approach that has been used chiefly in business applications, is used in medical diagnosis in this case. This paper is about a multi-agent system named “Distributed Ayurvedic Diagnosis and Therapy System for Hridroga using Agents” (DADTSHUA). It describes the architecture of the DADTSHUA model .This system is using mobile agents and ontology for passing data through the network. Due to this, transport delay can be minimized. It is a system which will be very helpful for the beginning physicians to eliminate his ambiguity in diagnosis and therapy. The system is implemented using Java Agent DEvelopment framework (JADE), which is a java-complaint mobile agent platform from TILab.

An Online Character Recognition System to Convert Grantha Script to Malayalam

This paper presents a novel approach to recognize Grantha, an ancient script in South India and converting it to Malayalam, a prevalent language in South India using online character recognition mechanism. The motivation behind this work owes its credit to (i) developing a mechanism to recognize Grantha script in this modern world and (ii) affirming the strong connection among Grantha and Malayalam. A framework for the recognition of Grantha script using online character recognition is designed and implemented. The features extracted from the Grantha script comprises mainly of time-domain features based on writing direction and curvature. The recognized characters are mapped to corresponding Malayalam characters. The framework was tested on a bed of medium length manuscripts containing 9-12 sample lines and printed pages of a book titled Soundarya Lahari writtenin Grantha by Sri Adi Shankara to recognize the words and sentences. The manuscript recognition rates with the system are for Grantha as 92.11%, Old Malayalam 90.82% and for new Malayalam script 89.56%. The recognition rates of pages of the printed book are for Grantha as 96.16%, Old Malayalam script 95.22% and new Malayalam script as 92.32% respectively. These results show the efficiency of the developed system

Automated Transcription System for Malayalam Language

Malayalam is one of the 22 scheduled languages in India with more than 130 million speakers. This paper presents a report on the development of a speaker independent, continuous transcription system for Malayalam. The system employs Hidden Markov Model (HMM) for acoustic modeling and Mel Frequency Cepstral Coefficient (MFCC) for feature extraction. It is trained with 21 male and female speakers in the age group ranging from 20 to 40 years. The system obtained a word recognition accuracy of 87.4% and a sentence recognition accuracy of 84%, when tested with a set of continuous speech data.

Content Based Image Retrieval System for Malayalam Handwritten Characters

Content Based Image Retrieval is one of the prominent areas in Computer Vision and Image Processing. Recognition of handwritten characters has been a popular area of research for many years and still remains an open problem. The proposed system uses visual image queries for retrieving similar images from database of Malayalam handwritten characters. Local Binary Pattern (LBP) descriptors of the query images are extracted and those features are compared with the features of the images in database for retrieving desired characters. This system with local binary pattern gives excellent retrieval performance

Developing a Child Friendly Text-to-Speech System

This paper discusses the implementation details of a child friendly, good quality, English text-to-speech (TTS) system that is phoneme-based, concatenative, easy to set up and use with little memory. Direct waveform concatenation and linear prediction coding (LPC) are used. Most existing TTS systems are unit-selection based, which use standard speech databases available in neutral adult voices.Here reduced memory is achieved by the concatenation of phonemes and by replacing phonetic wave files with their LPC coefficients. Linguistic analysis was used to reduce the algorithmic complexity instead of signal processing techniques. Sufficient degree of customization and generalization catering to the needs of the child user had been included through the provision for vocabulary and voice selection to suit the requisites of the child. Prosody had also been incorporated. This inexpensive TTS systemwas implemented inMATLAB, with the synthesis presented by means of a graphical user interface (GUI), thus making it child friendly. This can be used not only as an interesting language learning aid for the normal child but it also serves as a speech aid to the vocally disabled child. The quality of the synthesized speech was evaluated using the mean opinion score (MOS).

Design of a Knowledge-Based Geographic Information System

A GIS has been designed with limited Functionalities; but with a novel approach in Aits design. The spatial data model adopted in the design of KBGIS is the unlinked vector model. Each map entity is encoded separately in vector fonn, without referencing any of its neighbouring entities. Spatial relations, in other words, are not encoded. This approach is adequate for routine analysis of geographic data represented on a planar map, and their display (Pages 105-106). Even though spatial relations are not encoded explicitly, they can be extracted through the specially designed queries. This work was undertaken as an experiment to study the feasibility of developing a GIS using a knowledge base in place of a relational database. The source of input spatial data was accurate sheet maps that were manually digitised. Each identifiable geographic primitive was represented as a distinct object, with its spatial properties and attributes defined. Composite spatial objects, made up of primitive objects, were formulated, based on production rules defining such compositions. The facts and rules were then organised into a production system, using OPS5