Domain Adaptation in remote sensing: increasing the portability of land-cover classifiers

Among the types of remote sensing acquisitions, optical images are certainly one of the most widely relied upon data sources for Earth observation. They provide detailed measurements of the electromagnetic radiation reflected or emitted by each pixel in the scene. Through a process termed supervised land-cover classification, this allows to automatically yet accurately distinguish objects at the surface of our planet. In this respect, when producing a land-cover map of the surveyed area, the availability of training examples representative of each thematic class is crucial for the success of the classification procedure. However, in real applications, due to several constraints on the sample collection process, labeled pixels are usually scarce. When analyzing an image for which those key samples are unavailable, a viable solution consists in resorting to the ground truth data of other previously acquired images. This option is attractive but several factors such as atmospheric, ground and acquisition conditions can cause radiometric differences between the images, hindering therefore the transfer of knowledge from one image to another. The goal of this Thesis is to supply remote sensing image analysts with suitable processing techniques to ensure a robust portability of the classification models across different images. The ultimate purpose is to map the land-cover classes over large spatial and temporal extents with minimal ground information. To overcome, or simply quantify, the observed shifts in the statistical distribution of the spectra of the materials, we study four approaches issued from the field of machine learning. First, we propose a strategy to intelligently sample the image of interest to collect the labels only in correspondence of the most useful pixels. This iterative routine is based on a constant evaluation of the pertinence to the new image of the initial training data actually belonging to a different image. Second, an approach to reduce the radiometric differences among the images by projecting the respective pixels in a common new data space is presented. We analyze a kernel-based feature extraction framework suited for such problems, showing that, after this relative normalization, the cross-image generalization abilities of a classifier are highly increased. Third, we test a new data-driven measure of distance between probability distributions to assess the distortions caused by differences in the acquisition geometry affecting series of multi-angle images. Also, we gauge the portability of classification models through the sequences. In both exercises, the efficacy of classic physically- and statistically-based normalization methods is discussed. Finally, we explore a new family of approaches based on sparse representations of the samples to reciprocally convert the data space of two images. The projection function bridging the images allows a synthesis of new pixels with more similar characteristics ultimately facilitating the land-cover mapping across images.

Domain adaptation per classificazione di point cloud mediante pseudo-annotazioni

La classificazione di dati geometrici 3D come point cloud è un tema emergente nell'ambito della visione artificiale in quanto trova applicazione in molteplici contesti di guida autonoma, robotica e realtà aumentata. Sebbene nel mercato siano presenti una grande quantità di sensori in grado di ottenere scansioni reali, la loro annotazione costituisce un collo di bottiglia per la generazione di dataset. Per sopperire al problema si ricorre spesso alla domain adaptation sfruttando dati sintetici annotati. Questo elaborato si pone come obiettivo l'analisi e l'implementazione di metodi di domain adaptation per classificazione di point cloud mediante pseudo-labels. In particolare, sono stati condotti esperimenti all'interno del framework RefRec valutando la possibilità di sostituire nuove architetture di deep learning al modello preesistente. Tra queste, Transformer con mascheramento dell'input ha raggiunto risultati superiori allo stato dell'arte nell'adattamento da dati sintetici a reali (ModelNet->ScanNet) esaminato in questa tesi.

Mining and exploiting domain-specific corpora in the PANACEA platform

The objective of the PANACEA ICT-2007.2.2 EU project is to build a platform that automates the stages involved in the acquisition,production, updating and maintenance of the large language resources required by, among others, MT systems. The development of a Corpus Acquisition Component (CAC) for extracting monolingual and bilingual data from the web is one of the most innovative building blocks of PANACEA. The CAC, which is the first stage in the PANACEA pipeline for building Language Resources, adopts an efficient and distributed methodology to crawl for web documents with rich textual content in specific languages and predefined domains. The CAC includes modules that can acquire parallel data from sites with in-domain content available in more than one language. In order to extrinsically evaluate the CAC methodology, we have conducted several experiments that used crawled parallel corpora for the identification and extraction of parallel sentences using sentence alignment. The corpora were then successfully used for domain adaptation of Machine Translation Systems.

Apprentissage des réseaux de neurones profonds et applications en traitement automatique de la langue naturelle

En apprentissage automatique, domaine qui consiste à utiliser des données pour apprendre une solution aux problèmes que nous voulons confier à la machine, le modèle des Réseaux de Neurones Artificiels (ANN) est un outil précieux. Il a été inventé voilà maintenant près de soixante ans, et pourtant, il est encore de nos jours le sujet d'une recherche active. Récemment, avec l'apprentissage profond, il a en effet permis d'améliorer l'état de l'art dans de nombreux champs d'applications comme la vision par ordinateur, le traitement de la parole et le traitement des langues naturelles. La quantité toujours grandissante de données disponibles et les améliorations du matériel informatique ont permis de faciliter l'apprentissage de modèles à haute capacité comme les ANNs profonds. Cependant, des difficultés inhérentes à l'entraînement de tels modèles, comme les minima locaux, ont encore un impact important. L'apprentissage profond vise donc à trouver des solutions, en régularisant ou en facilitant l'optimisation. Le pré-entraînnement non-supervisé, ou la technique du ``Dropout'', en sont des exemples. Les deux premiers travaux présentés dans cette thèse suivent cette ligne de recherche. Le premier étudie les problèmes de gradients diminuants/explosants dans les architectures profondes. Il montre que des choix simples, comme la fonction d'activation ou l'initialisation des poids du réseaux, ont une grande influence. Nous proposons l'initialisation normalisée pour faciliter l'apprentissage. Le second se focalise sur le choix de la fonction d'activation et présente le rectifieur, ou unité rectificatrice linéaire. Cette étude a été la première à mettre l'accent sur les fonctions d'activations linéaires par morceaux pour les réseaux de neurones profonds en apprentissage supervisé. Aujourd'hui, ce type de fonction d'activation est une composante essentielle des réseaux de neurones profonds. Les deux derniers travaux présentés se concentrent sur les applications des ANNs en traitement des langues naturelles. Le premier aborde le sujet de l'adaptation de domaine pour l'analyse de sentiment, en utilisant des Auto-Encodeurs Débruitants. Celui-ci est encore l'état de l'art de nos jours. Le second traite de l'apprentissage de données multi-relationnelles avec un modèle à base d'énergie, pouvant être utilisé pour la tâche de désambiguation de sens.

Deep convolutional neural networks for semantic video object segmentation

In this thesis, we propose to infer pixel-level labelling in video by utilising only object category information, exploiting the intrinsic structure of video data. Our motivation is the observation that image-level labels are much more easily to be acquired than pixel-level labels, and it is natural to find a link between the image level recognition and pixel level classification in video data, which would transfer learned recognition models from one domain to the other one. To this end, this thesis proposes two domain adaptation approaches to adapt the deep convolutional neural network (CNN) image recognition model trained from labelled image data to the target domain exploiting both semantic evidence learned from CNN, and the intrinsic structures of unlabelled video data. Our proposed approaches explicitly model and compensate for the domain adaptation from the source domain to the target domain which in turn underpins a robust semantic object segmentation method for natural videos. We demonstrate the superior performance of our methods by presenting extensive evaluations on challenging datasets comparing with the state-of-the-art methods.

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.

Hierarchical Multi-Label Text Classification in a Low-Resource Setting

In this thesis we address a multi-label hierarchical text classification problem in a low-resource setting and explore different approaches to identify the best one for our case. The goal is to train a model that classifies English school exercises according to a hierarchical taxonomy with few labeled data. The experiments made in this work employ different machine learning models and text representation techniques: CatBoost with tf-idf features, classifiers based on pre-trained models (mBERT, LASER), and SetFit, a framework for few-shot text classification. SetFit proved to be the most promising approach, achieving better performance when during training only a few labeled examples per class are available. However, this thesis does not consider all the hierarchical taxonomy, but only the first two levels: to address classification with the classes at the third level further experiments should be carried out, exploring methods for zero-shot text classification, data augmentation, and strategies to exploit the hierarchical structure of the taxonomy during training.

New genes of Xanthomonas citri subsp citri involved in pathogenesis and adaptation revealed by a transposon-based mutant library

Background: Citrus canker is a disease caused by the phytopathogens Xanthomonas citri subsp. citri, Xanthomonas fuscans subsp. aurantifolli and Xanthomonas alfalfae subsp. citrumelonis. The first of the three species, which causes citrus bacterial canker type A, is the most widely spread and severe, attacking all citrus species. In Brazil, this species is the most important, being found in practically all areas where citrus canker has been detected. Like most phytobacterioses, there is no efficient way to control citrus canker. Considering the importance of the disease worldwide, investigation is needed to accurately detect which genes are related to the pathogen-host adaptation process and which are associated with pathogenesis. Results: Through transposon insertion mutagenesis, 10,000 mutants of Xanthomonas citri subsp. citri strain 306 (Xcc) were obtained, and 3,300 were inoculated in Rangpur lime (Citrus limonia) leaves. Their ability to cause citrus canker was analyzed every 3 days until 21 days after inoculation; a set of 44 mutants showed altered virulence, with 8 presenting a complete loss of causing citrus canker symptoms. Sequencing of the insertion site in all 44 mutants revealed that 35 different ORFs were hit, since some ORFs were hit in more than one mutant, with mutants for the same ORF presenting the same phenotype. An analysis of these ORFs showed that some encoded genes were previously known as related to pathogenicity in phytobacteria and, more interestingly, revealed new genes never implicated with Xanthomonas pathogenicity before, including hypothetical ORFs. Among the 8 mutants with no canker symptoms are the hrpB4 and hrpX genes, two genes that belong to type III secretion system (TTSS), two hypothetical ORFS and, surprisingly, the htrA gene, a gene reported as involved with the virulence process in animal-pathogenic bacteria but not described as involved in phytobacteria virulence. Nucleic acid hybridization using labeled cDNA probes showed that some of the mutated genes are differentially expressed when the bacterium is grown in citrus leaves. Finally, comparative genomic analysis revealed that 5 mutated ORFs are in new putative pathogenicity islands. Conclusion: The identification of these new genes related with Xcc infection and virulence is a great step towards the understanding of plant-pathogen interactions and could allow the development of strategies to control citrus canker.

Adaptive deblocking filter for transform domain Wyner-Ziv video coding

Wyner-Ziv (WZ) video coding is a particular case of distributed video coding, the recent video coding paradigm based on the Slepian-Wolf and Wyner-Ziv theorems that exploits the source correlation at the decoder and not at the encoder as in predictive video coding. Although many improvements have been done over the last years, the performance of the state-of-the-art WZ video codecs still did not reach the performance of state-of-the-art predictive video codecs, especially for high and complex motion video content. This is also true in terms of subjective image quality mainly because of a considerable amount of blocking artefacts present in the decoded WZ video frames. This paper proposes an adaptive deblocking filter to improve both the subjective and objective qualities of the WZ frames in a transform domain WZ video codec. The proposed filter is an adaptation of the advanced deblocking filter defined in the H.264/AVC (advanced video coding) standard to a WZ video codec. The results obtained confirm the subjective quality improvement and objective quality gains that can go up to 0.63 dB in the overall for sequences with high motion content when large group of pictures are used.

Portuguese version of the Tilburg Frailty Indicator: Transcultural adaptation and psychometric validation

Artigo científico disponível actualmente em Early View (Online Version of Record published before inclusion in an issue)

Providing Adaptation for an Adaptive Hypermedia System

O ensino à distância cresceu consideravelmente nos últimos anos e a tendência é para que continue a crescer em anos vindouros. No entanto, enquanto que a maioria das plataformas de ensino à distância utilizam a mesma abordagem de ensino para todos os utilizadores, os estudantes que as usam são na realidade pessoas de diferentes culturas, locais, idades e géneros, e que possuem diferentes níveis de educação. Ao contrário do ensino à distância tradicional, os sistemas de hipermédia adaptativa educacional adaptam interface, apresentação de conteúdos e navegação, entre outros, às características, necessidades e interesses específicos de diferentes utilizadores. Apesar da investigação na área de sistemas de hipermédia adaptativa já estar bastante desenvolvida, é necessário efetuar mais desenvolvimento e experimentação de modo a determinar quais são os aspetos mais eficazes destes sistemas e avaliar o seu sucesso. A Plataforma de Aprendizagem Colaborativa da Matemática (PCMAT) é um sistema de hipermédia adaptativa educacional com uma abordagem construtivista, que foi desenvolvido com o objetivo de contribuir para a investigação na área de sistemas de hipermédia adaptativa. A plataforma avalia o conhecimento do utilizador e apresenta conteúdos e atividades adaptadas às características e estilo de aprendizagem dominante de estudantes de matemática do segundo ciclo. O desenvolvimento do PCMAT tem também o propósito de auxiliar os alunos Portugueses com a aprendizagem da matemática. De acordo com o estudo PISA 2012 da OCDE [OECD, 2014], o desempenho dos alunos Portugueses na área da matemática melhorou em relação à edição anterior do estudo, mas os resultados obtidos permanecem abaixo da média da OCDE. Por este motivo, uma das finalidades deste projeto é desenvolver um sistema de hipermédia adaptativa que, ao adequar o ensino da matemática às necessidades específicas de cada aluno, os assista com a aquisição de conhecimento. A adaptação é efetuada pelo sistema usando a informação constante no modelo do utilizador para definir um grafo de conceitos do domínio específico. Este grafo é adaptado do modelo do domínio e utilizado para dar resposta às necessidades particulares de cada aluno. Embora a trajetória inicial seja definida pelo professor, o percurso percorrido no grafo por cada aluno é determinado pela sua interação com o sistema, usando para o efeito a representação do conhecimento do aluno e outras características disponíveis no modelo do utilizador, assim como avaliação progressiva. A adaptação é conseguida através de alterações na apresentação de conteúdos e na estrutura e anotações das hiperligações. A apresentação de conteúdos é alterada mostrando ou ocultando cada um dos vários fragmentos que compõe as páginas dum curso. Estes fragmentos são compostos por diferentes objetos de aprendizagem, tais como exercícios, figuras, diagramas, etc. As mudanças efetuadas na estrutura e anotações das hiperligações têm o objetivo de guiar o estudante, apontando-o na direção do conhecimento mais relevante e mantendo-o afastado de informação inadequada. A escolha de objectos de aprendizagem adequados às características particulares de cada aluno é um aspecto essencial do modelo de adaptação do PCMAT. A plataforma inclui para esse propósito um módulo responsável pela recomendação de objectos de aprendizagem, e um módulo para a pesquisa e recuperação dos mesmos. O módulo de recomendação utiliza lógica Fuzzy para converter determinados atributos do aluno num conjunto de parâmetros que caracterizam o objecto de aprendizagem que idealmente deveria ser apresentado ao aluno. Uma vez que o objecto “ideal” poderá não existir no repositório de objectos de aprendizagem do sistema, esses parâmetros são utilizados pelo módulo de pesquisa e recuperação para procurar e devolver ao módulo de recomendação uma lista com os objectos que mais se assemelham ao objecto “ideal”. A pesquisa é feita numa árvore k-d usando o algoritmo k-vizinhos mais próximos. O modelo de recomendação utiliza a lista devolvida pelo módulo de pesquisa e recuperação para seleccionar o objecto de aprendizagem mais apropriado para o aluno e processa-o para inclusão numa das páginas Web do curso. O presente documento descreve o trabalho desenvolvido no âmbito do projeto PCMAT (PTDS/CED/108339/2008), dando relevância à adaptação de conteúdos.

Dynamic adaptation of interaction models for stateful web services

Dissertação para obtenção do Grau de Mestre em Engenharia Informática