Diese Diplomarbeit führt kurz in das Thema Information Retrieval mit den Schwerpunkten Evaluation und Evaluationskampagnen ein. Im Anschluss wird anhand der Nachteile eines vorhandenen Retrieval Systems ein neues Retrieval Framework zur experimentellen Evaluation von Ansätzen aus dem Information Retrieval entworfen und umgesetzt. Die Komponenten des Frameworks sind dabei so abstrakt angelegt, dass verschiedene, bestehende Retrieval Systeme, wie zum Beispiel Apache Lucene oder Terrier, integriert werden können. Anhand einer Referenzimplementierung für den ImageCLEF Photographic Retrieval Task des ImageCLEF Tracks des Cross Language Evaluation Forums wird die Funktionsfähigkeit des Frameworks überprüft und bestätigt.

Automotive technologies and fully autonomous driving have seen a tremendous growth in recent times and have benefitted from extensive deep learning research. State-of-the-art deep learning methods are largely supervised and require labelled data for training. However, the annotation process for image data is time-consuming and costly in terms of human efforts. It is of interest to find informative samples for labelling by Content Based Image Retrieval (CBIR). Generally, a CBIR method takes a query image as input and returns a set of images that are semantically similar to the query image. The image retrieval is achieved by transforming images to feature representations in a latent space, where it is possible to reason about image similarity in terms of image content. In this thesis, a self-supervised method is developed to learn feature representations of road scenes images. The self-supervised method learns feature representations for images by adapting intermediate convolutional features from an existing deep Convolutional Neural Network (CNN). A contrastive approach based on Noise Contrastive Estimation (NCE) is used to train the feature learning model. For complex images like road scenes where mutiple image aspects can occur simultaneously, it is important to embed all the salient image aspects in the feature representation. To achieve this, the output feature representation is obtained as an ensemble of feature embeddings which are learned by focusing on different image aspects. An attention mechanism is incorporated to encourage each ensemble member to focus on different image aspects. For comparison, a self-supervised model without attention is considered and a simple dimensionality reduction approach using SVD is treated as the baseline. The methods are evaluated on nine different evaluation datasets using CBIR performance metrics. The datasets correspond to different image aspects and concern the images at different spatial levels - global, semi-global and local. The feature representations learned by self-supervised methods are shown to perform better than the SVD approach. Taking into account that no labelled data is required for training, learning representations for road scenes images using self-supervised methods appear to be a promising direction. Usage of multiple query images to emphasize a query intention is investigated and a clear improvement in CBIR performance is observed. It is inconclusive whether the addition of an attentive mechanism impacts CBIR performance. The attention method shows some positive signs based on qualitative analysis and also performs better than other methods for one of the evaluation datasets containing a local aspect. This method for learning feature representations is promising but requires further research involving more diverse and complex image aspects.

Traditionally, local shape descriptors or color and texture based descriptors are used to describe the content of images. Although, these solutions achieving good results, they are not able to distinguish scenes that contain objects with the same colors, but with a different spatial organization or do not supports partial matching. In this work we focus on a particular case of the partial matching that is to find individual objects in images that contain various objects. Since the color is one of the most visually distinguishable properties, we propose a new descriptor based only on color able to find pictures of objects that are contained in other images. Although our descriptor has shown better results when compared to related works, this new color descriptor is not able to discriminate objects topologically different but having the same colors. To overcome this problem, we also propose a new approach to the partial matching of images that combine color and topological features on a single descriptor. This new descriptor, first performs a simplification process of the original image, which identifies the color regions that make up the image. Then, we represent the spatial information among the color regions using a topological graph, where vertices represent the color regions and the edges represent the spatial connections between them. To calculate the descriptor from this graph representation, we use the spectral theory of graphs, avoiding the need to make a direct comparison between graphs. To support the partial matching, we propose a decomposition of the main graph into several subgraphs, and also calculate descriptors for these subgraphs. / Tradicionalmente, descritores de forma, ou descritores baseados em cor e textura, são utilizados para descrever o conteúdo visual das imagens. Embora essas abordagens apresentem bons resultados, elas não são capazes de diferenciar adequadamente imagens que contêm objetos com as mesmas cores, mas com organização espacial diferente ou não suportam a pesquisa parcial de imagens. Neste trabalho focamos em um caso particular da pesquisa parcial de imagens, que é encontrar objetos em imagens que contenham vários objetos, não deixando de lado a pesquisa total (encontrar imagens similares à original). Dado que a cor é uma das propriedades visuais mais discriminativas, propomos um novo descritor baseado somente em cor capaz de encontrar imagens de objetos que estão contidos em outras imagens. Embora tenha apresentado melhores resultados quando comparado a trabalhos correlatos, esse novo descritor de cor não é capaz de discriminar objetos topologicamente diferentes mas que possuam as mesmas cores. Com o intuito de resolver esse problema, também propomos uma nova abordagem para a recuperação parcial de imagens que combina características topológicas e de cor em um único descritor. Esse novo descritor primeiramente realiza um processo de simplificação da imagem original, onde são identificadas as regiões de cor que compõem a imagem. Após esse processo de simplificação, a organização espacial das regiões de cor previamente identificadas é representada por meio de um grafo topológico, onde os vértices representam as regiões de cor e as arestas representam as conexões entre essas regiões. O descritor topológico é então calculado a partir do grafo de topologia utilizando a teoria espectral de grafos, evitando a necessidade de se realizar uma comparação direta entre grafos. Para suportar a pesquisa parcial de imagens, é realizada uma decomposição do grafo principal em diversos subgrafos. / Mestre em Ciência da Computação

While historically digital heritage libraries were first powered in image mode, they quickly took advantage of OCR technology to index printed collections and consequently improve the scope and performance of the information retrieval services offered to users. But the access to iconographic resources has not progressed in the same way, and the latter remain in the shadows: manual incomplete and heterogeneous indexation, data silos by iconographic genre. Today, however, it would be possible to make better use of these resources, especially by exploiting the enormous volumes of OCR produced during the last two decades, and thus valorize these engravings, drawings, photographs, maps, etc. for their own value but also as an attractive entry point into the collections, supporting discovery and serenpidity from document to document and collection to collection. This article presents an ETL (extract-transform-load) approach to this need, that aims to: Identify and extract iconography wherever it may be found, in image collections but also in printed materials (dailies, magazines, monographies); Transform, harmonize and enrich the image descriptive metadata (in particular with machine learning classification tools); Load it all into a web app dedicated to image retrieval. The approach is pragmatically dual, since it involves leveraging existing digital resources and (virtually) on-the-shelf technologies. / Si historiquement, les bibliothèques numériques patrimoniales furent d’abord alimentées par des images, elles profitèrent rapidement de la technologie OCR pour indexer les collections imprimées afin d’améliorer périmètre et performance du service de recherche d’information offert aux utilisateurs. Mais l’accès aux ressources iconographiques n’a pas connu les mêmes progrès et ces dernières demeurent dans l’ombre : indexation manuelle lacunaire, hétérogène et non viable à grande échelle ; silos documentaires par genre iconographique ; recherche par le contenu (CBIR, content-based image retrieval) encore peu opérationnelle sur les collections patrimoniales. Aujourd’hui, il serait pourtant possible de mieux valoriser ces ressources, en particulier en exploitant les énormes volumes d’OCR produits durant les deux dernières décennies (tant comme descripteur textuel que pour l’identification automatique des illustrations imprimées). Et ainsi mettre en valeur ces gravures, dessins, photographies, cartes, etc. pour leur valeur propre mais aussi comme point d’entrée dans les collections, en favorisant découverte et rebond de document en document, de collection à collection. Cet article décrit une approche ETL (extract-transform-load) appliquée aux images d’une bibliothèque numérique à vocation encyclopédique : identifier et extraire l’iconographie partout où elle se trouve (dans les collections image mais aussi dans les imprimés : presse, revue, monographie) ; transformer, harmoniser et enrichir ses métadonnées descriptives grâce à des techniques d’apprentissage machine – machine learning – pour la classification et l’indexation automatiques ; charger ces données dans une application web dédiée à la recherche iconographique (ou dans d’autres services de la bibliothèque). Approche qualifiée de pragmatique à double titre, puisqu’il s’agit de valoriser des ressources numériques existantes et de mettre à profit des technologies (quasiment) mâtures.