Abstract: Shape representation and registration are two important problems in computer vision and graphics. Representing the given cloud of points through an implicit function provides a higher level information describing the data. This representation can be more compact more robust to noise and outliers, hence it can be exploited in different computer vision application. In the first part of this thesis implicit shape representations, including both implicit B-spline and polynomial, are tackled. First, an approximation of a geometric distance is proposed to measure the closeness of the given cloud of points and the implicit surface. The analysis of the proposed distance shows an accurate estimation with smooth behavior. The distance by itself is used in a RANSAC based quadratic fitting method. Moreover, since the gradient information of the distance with respect to the surface parameters can be analytically computed, it is used in Levenberg-Marquadt algorithm to refine the surface parameters. In a different approach, an algebraic fitting method is used to represent an object through implicit B-splines. The outcome is a smooth flexible surface and can be represented in different levels from coarse to fine. This property has been exploited to solve the registration problem in the second part of the thesis. In the proposed registration technique the model set is replaced with an implicit representation provided in the first part; then, the point-to-point registration is converted to a point-to-model one in a higher level. This registration error can benefit from different distance estimations to speed up the registration process even without need of correspondence search. Finally, the non-rigid registration problem is tackled through a quadratic distance approximation that is based on the curvature information of the model set. This approximation is used in a free form deformation model to update its control lattice. Then it is shown how an accurate distance approximation can benefit non-rigid registration problems.

Abstract: Handwritten word-spotting (HWS) is the pattern analysis task that consists in finding keywords in handwritten document images. So far, HWS has been applied mostly to historical documents in order to build search engines for such image collections. This thesis addresses the problem of word-spotting for detecting important keywords in business documents. This is a first step towards the process of automatic routing of correspondence based on content.
However, the application of traditional HWS techniques fails for this type of documents. As opposed to historical documents, real business documents present a very high variability in terms of writing styles, spontaneous writing, crossed-out words, spelling mistakes, etc. The main goal of this thesis is the development of pattern recognition techniques that lead to a high-performance HWS system for this challenging type of data.
We develop a statistical framework in which word models are expressed in terms of hidden Markov models and the a priori information is encoded in a universal vocabulary of Gaussian codewords. This systems leads to a very robust performance in word-spotting task. We also find that by constraining the word models to the universal vocabulary, the a priori information of the problem of interest can be exploited for developing new contributions. These include a novel writer adaptation method, a system for searching handwritten words by generating typed text images, and a novel model-based similarity between feature vector sequences.

Abstract: Cardiovascular Diseases, specially those affecting the Left Ventricle (LV), are the leading cause of death in developed countries with approximately a 30% of all global deaths. In order to address this public health concern, physicians focus on diagnosis and therapy planning. On one hand, early and accurate detection of Regional Wall Motion Abnormalities (RWMA) significantly contributes to a quick diagnosis and prevents the patient to reach more severe stages. On the other hand, a thouroughly knowledge of the normal gross anatomy of the LV, as well as, the distribution of its muscular fibers is crucial for designing specific interventions and therapies (such as pacemaker implanction). Statistical models obtained from the analysis of different imaging modalities allow the computation of the normal ranges of variation within a given population. Normality models are a valuable tool for the definition of objective criterions quantifying the degree of (anomalous) deviation of the LV function and anatomy for a given subject. The creation of statistical models involve addressing three main issues: extraction of data from images, definition of a common domain for comparison of data across patients and designing appropriate statistical analysis schemes. In this PhD thesis we present generic image processing tools for the creation of statistical models of the LV anatomy and function. On one hand, we use differential geometry concepts to define a computational framework (the Normalized Parametric Domain, NPD) suitable for the comparison and fusion of several clinical scores obtained over the LV. On the other hand, we present a variational approach (the Harmonic Phase Flow, HPF) for the estimation of myocardial motion that provides dense and continuous vector fields without overestimating motion at injured areas. These tools are used for the creation of statistical models. Regarding anatomy, we obtain an atlas jointly modelling, both, LV gross anatomy and fiber architecture. Regarding function, we compute normality patterns of scores characterizing the (global and local) LV function and explore, for the first time, the configuration of local scores better suited for RWMA detection.

Abstract: Understanding the content of the images is one of the greatest challenges of computer vision. Recognition of objects appearing in images, identifying and interpreting their actions are the main purposes of Image Understanding. This thesis seeks to identify what is present in a picture by categorizing and locating all the objects in the scene.
Images are composed by pixels, and one possibility consists of assigning to each pixel an object category, which is commonly known as semantic segmentation. By incorporating information as a contextual cue, we are able to resolve the ambiguity within categories at the pixel-level. We propose three levels of scale in order to resolve such ambiguity.
Another possibility to represent the objects is the object detection task. In this case, the aim is to recognize and localize the whole object by accurately placing a bounding box around it. We present two new approaches. The first one is focused on improving the object representation of deformable part models with the concept of factorized appearances. The second approach addresses the issue of reducing the computational cost for multi-class recognition. The results given have been validated on several commonly used datasets, reaching international recognition and state-of-the-art within the field

Abstract: In the current work, we identify several problems of current tracking systems. The lack of large-scale labeled datasets hampers the usage of deep learning, especially end-to-end training, for tracking in TIR images. Therefore, many methods for tracking on TIR data are still based on hand-crafted features. This situation also happens in multi-modal tracking, e.g. RGB-T tracking. Another reason, which hampers the development of RGB-T tracking, is that there exists little research on the fusion mechanisms for combining information from RGB and TIR modalities. One of the crucial components of most trackers is the update module. For the currently existing end-to-end tracking architecture, e.g, Siamese trackers, the online model update is still not taken into consideration at the training stage. They use no-update or a linear update strategy during the inference stage. While such a hand-crafted approach to updating has led to improved results, its simplicity limits the potential gain likely to be obtained by learning to update.
To address the data-scarcity for TIR and RGB-T tracking, we use image-to-image translation to generate a large-scale synthetic TIR dataset. This dataset allows us to perform end-to-end training for TIR tracking. Furthermore, we investigate several fusion mechanisms for RGB-T tracking. The multi-modal trackers are also trained in an end-to-end manner on the synthetic data. To improve the standard online update, we pose the updating step as an optimization problem which can be solved by training a neural network. Our approach thereby reduces the hand-crafted components in the tracking pipeline and sets a further step in the direction of a complete end-to-end trained tracking network which also considers updating during optimization.

Abstract: Fine-grained recognition, i.e. identifying similar subcategories of the same superclass, is central to human activity. Recognizing a friend, finding bacteria in microscopic imagery, or discovering a new kind of galaxy, are just but few examples. However, fine-grained image recognition is still a challenging computer vision task since the differences between two images of the same category can overwhelm the differences between two images of different fine-grained categories. In this regime, where the difference between two categories resides on subtle input changes, excessively invariant CNNs discard those details that help to discriminate between categories and focus on more obvious changes, yielding poor classification performance.
On the other hand, CNNs with too much capacity tend to memorize instance-specific details, thus causing overfitting. In this thesis,motivated by the
potential impact of automatic fine-grained image recognition, we tackle the previous challenges and demonstrate that proper alignment of the inputs, multiple levels of attention, regularization, and explicitmodeling of the output space, results inmore accurate fine-grained recognitionmodels, that generalize better, and are more robust to intra-class variation. Concretely, we study the different stages of the neural network pipeline: input pre-processing, attention to regions, feature activations, and the label space. In each stage, we address different issues that hinder the recognition performance on various fine-grained tasks, and devise solutions in each chapter: i)We deal with the sensitivity to input alignment on fine-grained human facial motion such as pain. ii) We introduce an attention mechanism to allow CNNs to choose and process in detail the most discriminate regions of the image. iii)We further extend attention mechanisms to act on the network activations,
thus allowing them to correct their predictions by looking back at certain
regions, at different levels of abstraction. iv) We propose a regularization loss to prevent high-capacity neural networks to memorize instance details by means of almost-identical feature detectors. v)We finally study the advantages of explicitly modeling the output space within the error-correcting framework. As a result, in this thesis we demonstrate that attention and regularization seem promising directions to overcome the problems of fine-grained image recognition, as well as proper treatment of the input and the output space.

Abstract: Artificial intelligence is innovating the fashion industry by proposing new applications and solutions to the problems encountered by researchers and engineers working in the industry. In this thesis, we address three of these problems. In the first part of the thesis, we tackle the problem of multi-label image classification which is very related to fashion attribute recognition. In the second part of the thesis, we address two problems that are specific to fashion. Firstly, we address the problem of main product detection which is the task of associating correct image parts (e.g. bounding boxes) with the fashion product being sold. Secondly, we address the problem of color naming for multicolored fashion items. The task of multi-label image classification consists in assigning various concepts such as objects or attributes to images. Usually, there are dependencies that can be learned between the concepts to capture label correlations (chair and table classes are more likely to co-exist than chair and giraffe).
If we treat the multi-label image classification problem as an orderless set prediction problem, we can exploit recurrent neural networks (RNN) to capture label correlations. However, RNNs are trained to predict ordered sequences of tokens, so if the order of the predicted sequence is different than the order of the ground truth sequence, there will be penalization although the predictions are correct. Therefore, in the first part of the thesis, we propose an orderless loss function which will order the labels in the ground truth sequence dynamically in a way that the minimum loss is achieved. This results in a significant improvement of RNN models on multi-label image classification over the previous methods.
However, RNNs suffer from long term dependencies when the cardinality of set grows bigger. The decoding process might stop early if the current hidden state cannot find any object and outputs the termination token. This would cause the remaining classes not to be predicted and lower recall metric. Transformers can be used to avoid the long term dependency problem exploiting their selfattention modules that process sequential data simultaneously. Consequently, we propose a novel transformer model for multi-label image classification which surpasses the state-of-the-art results by a large margin.
In the second part of thesis, we focus on two fashion-specific problems. Main product detection is the task of associating image parts with the fashion product that is being sold, generally using associated textual metadata (product title or description). Normally, in fashion e-commerces, products are represented by multiple images where a person wears the product along with other fashion items. If all the fashion items in the images are marked with bounding boxes, we can use the textual metadata to decide which item is the main product. The initial work treated each of these images independently, discarding the fact that they all belong to the same product. In this thesis, we represent the bounding boxes from all the images as nodes in a fully connected graph. This allows the algorithm to learn relations between the nodes during training and take the entire context into account for the final decision. Our algorithm results in a significant improvement of the state-ofthe-art.
Moreover, we address the problem of color naming for multicolored fashion items, which is a challenging task due to the external factors such as illumination changes or objects that act as clutter. In the context of multi-label classification, the vaguely defined lines between the classes in the color space cause ambiguity. For example, a shade of blue which is very close to green might cause the model to incorrectly predict the color blue and green at the same time. Based on this, models trained for color naming are expected to recognize the colors and their quantities in both single colored and multicolored fashion items. Therefore, in this thesis, we propose a novel architecture with an additional head that explicitly estimates the number of colors in fashion items. This removes the ambiguity problem and results in better color naming performance.

Abstract: Unsupervised representation learning aims at finding methods that learn representations from data without annotation-based signals. Abstaining from annotations not only leads to economic benefits but may – and to some extent already does – result in advantages regarding the representation’s structure, robustness, and generalizability to different tasks. In the long run, unsupervised methods are expected to surpass their supervised counterparts due to the reduction of human intervention and the inherently more general setup that does not bias the optimization towards an objective originating from specific annotation-based signals. While major advantages of unsupervised representation learning have been recently observed in natural language processing, supervised methods still dominate in vision domains for most tasks. In this dissertation, we contribute to the field of unsupervised (visual) representation learning from three perspectives: (i) Learning representations: We design unsupervised, backpropagation-free Convolutional Self-Organizing Neural Networks (CSNNs) that utilize self-organization- and Hebbian-based learning rules to learn convolutional kernels and masks to achieve deeper backpropagation-free models. Thereby, we observe that backpropagation-based and -free methods can suffer from an objective function mismatch between the unsupervised pretext task and the target task. This mismatch can lead to performance decreases for the target task. (ii) Evaluating representations: We build upon the widely used (non-)linear evaluation protocol to define pretext- and target-objective-independent metrics for measuring the objective function mismatch. With these metrics, we evaluate various pretext and target tasks and disclose dependencies of the objective function mismatch concerning different parts of the training and model setup. (iii) Transferring representations: We contribute CARLANE, the first 3-way sim-to-real domain adaptation benchmark for 2D lane detection. We adopt several well-known unsupervised domain adaptation methods as baselines and propose a method based on prototypical cross-domain self-supervised learning. Finally, we focus on pixel-based unsupervised domain adaptation and contribute a content-consistent unpaired image-to-image translation method that utilizes masks, global and local discriminators, and similarity sampling to mitigate content inconsistencies, as well as feature-attentive denormalization to fuse content-based statistics into the generator stream. In addition, we propose the cKVD metric to incorporate class-specific content inconsistencies into perceptual metrics for measuring translation quality.