Abstract: The analysis and recognition of historical document images has attracted growing interest in the last years. Mass digitization and document image understanding allows the preservation, access and indexation of this artistic, cultural and technical heritage. The analysis of handwritten documents is an outstanding subfield. The main interest is not only the transcription of the document to a standard format, but also, the identification of the author of a document from a set of writers (namely writer identification).
Writer identification in handwritten text documents is an active area of study, however, the identification of the writer of graphical documents is still a challenge. The main objective of this thesis is the identification of the writer in old music scores, as an example of graphic documents. Concerning old music scores, many historical archives contain a huge number of sheets of musical compositions without information about the composer, and the research on this field could be helpful for musicologists.
The writer identification framework proposed in this thesis combines three different writer identification approaches, which are the main scientific contributions. The first one is based on symbol recognition methods. For this purpose, two novel symbol recognition methods are proposed for coping with the typical distortions in hand-drawn symbols. The second approach preprocesses the music score for obtaining music lines, and extracts information about the slant, width of the writing, connected components, contours and fractals. Finally, the third approach extracts global information by generating texture images from the music scores and extracting textural features (such as Gabor filters and co-occurence matrices).
The high identification rates obtained in the experimental results demonstrate the suitability of the proposed ensemble architecture. To the best of our knowledge, this work is the first contribution on writer identification from images containing graphical languages.

Abstract: Making Ground Autonomous Vehicles (GAVs) a reality as a service for the society is one of the major scientific and technological challenges of this century. The potential benefits of autonomous vehicles include reducing accidents, improving traffic congestion and better usage of road infrastructures, among others. These vehicles must operate in our cities, towns and highways, dealing with many different types of situations while respecting traffic rules and protecting human lives. GAVs are expected to deal with all types of scenarios and situations, coping with an uncertain and chaotic world.
Therefore, in order to fulfill these demanding requirements GAVs need to be endowed with the capability of understanding their surrounding at many different levels, by means of affordable sensors and artificial intelligence. This capacity to understand the surroundings and the current situation that the vehicle is involved in is called scene understanding. In this work we investigate novel techniques to bring scene understanding to autonomous vehicles by combining the use of cameras as the main source of information—due to their versatility and affordability—and algorithms based on computer vision and machine learning. We investigate different degrees of understanding of the scene, starting from basic geometric knowledge about where is the vehicle within the scene. A robust and efficient estimation of the vehicle location and pose with respect to a map is one of the most fundamental steps towards autonomous driving. We study this problem from the point of view of robustness and computational efficiency, proposing key insights to improve current solutions. Then we advance to higher levels of abstraction to discover what is in the scene, by recognizing and parsing all the elements present on a driving scene, such as roads, sidewalks, pedestrians, etc. We investigate this problem known as semantic segmentation, proposing new approaches to improve recognition accuracy and computational efficiency. We cover these points by focusing on key aspects such as: (i) how to leverage computation moving semantics to an offline process, (ii) how to train compact architectures based on deconvolutional networks to achieve their maximum potential, (iii) how to use virtual worlds in combination with domain adaptation to produce accurate models in a cost-effective fashion, and (iv) how to use transfer learning techniques to prepare models to new situations. We finally extend the previous level of knowledge enabling systems to reasoning about what has change in a scene with respect to a previous visit, which in return allows for efficient and cost-effective map updating.

Abstract: Deep convolutional neural networks (CNNs) have achieved superior performance in many visual recognition application, such as image classification, detection and segmentation. In this thesis we address two limitations of CNNs. Training deep CNNs requires huge amounts of labeled data, which is expensive and labor intensive to collect. Another limitation is that training CNNs in a continual learning setting is still an open research question. Catastrophic forgetting is very likely when adapting trained models to new environments or new tasks. Therefore, in this thesis, we aim to improve CNNs for applications with limited data and to adapt CNNs continually to new tasks.
Self-supervised learning leverages unlabelled data by introducing an auxiliary task for which data is abundantly available. In the first part of the thesis, we show how rankings can be used as a proxy self-supervised task for regression problems. Then we propose an efficient backpropagation technique for Siamese networks which prevents the redundant computation introduced by the multi-branch network architecture. In addition, we show that measuring network uncertainty on the self-supervised proxy task is a good measure of informativeness of unlabeled data. This can be used to drive an algorithm for active learning. We then apply our framework on two regression problems: Image Quality Assessment (IQA) and Crowd Counting. For both, we show how to automatically generate ranked image sets from unlabeled data. Our results show that networks trained to regress to the ground truth targets for labeled data and to simultaneously learn to rank unlabeled data obtain significantly better, state-of-the-art results. We further show that active learning using rankings can reduce labeling effort by up to 50\% for both IQA and crowd counting.
In the second part of the thesis, we propose two approaches to avoiding catastrophic forgetting in sequential task learning scenarios. The first approach is derived from Elastic Weight Consolidation, which uses a diagonal Fisher Information Matrix (FIM) to measure the importance of the parameters of the network. However the diagonal assumption is unrealistic. Therefore, we approximately diagonalize the FIM using a set of factorized rotation parameters. This leads to significantly better performance on continual learning of sequential tasks. For the second approach, we show that forgetting manifests differently at different layers in the network and propose a hybrid approach where distillation is used in the feature extractor and replay in the classifier via feature generation. Our method addresses the limitations of generative image replay and probability distillation (i.e. learning without forgetting) and can naturally aggregate new tasks in a single, well-calibrated classifier. Experiments confirm that our proposed approach outperforms the baselines and some start-of-the-art methods.

Abstract: Pattern recognition is the task that aims at distinguishing objects among different classes. When such a task wants to be solved in an automatic way a crucial step is how to formally represent such patterns to the computer. Based on the different representational formalisms, we may distinguish between statistical and structural pattern recognition. The former describes objects as a set of measurements arranged in the form of what is called a feature vector. The latter assumes that relations between parts of the underlying objects need to be explicitly represented and thus it uses relational structures such as graphs for encoding their inherent information. Vector spaces are a very flexible mathematical structure that has allowed to come up with several efficient ways for the analysis of patterns under the form of feature vectors. Nevertheless, such a representation cannot explicitly cope with binary relations between parts of the objects and it is restricted to measure the exact same number of features for each pattern under study regardless of their complexity. Graph-based representations present the contrary situation. They can easily adapt to the inherent complexity of the patterns but introduce a problem of high computational complexity, hindering the design of efficient tools to process and analyse patterns.
Solving this paradox is the main goal of this thesis. The ideal situation for solving pattern recognition problems would be to represent the patterns using relational structures such as graphs, and to be able to use the wealthy repository of data processing tools from the statistical pattern recognition domain. An elegant solution to this problem is to transform the graph domain into a vector domain where any processing algorithm can be applied. In other words, by mapping each graph to a point in a vector space we automatically get access to the rich set of algorithms from the statistical domain to be applied in the graph domain. Such methodology is called graph embedding.
In this thesis we propose to associate feature vectors to graphs in a simple and very efficient way by just putting attention on the labelling information that graphs store. In particular, we count frequencies of node labels and of edges between labels. Although their locality, these features are able to robustly represent structurally global properties of graphs, when considered together in the form of a vector. We initially deal with the case of discrete attributed graphs, where features are easy to compute. The continuous case is tackled as a natural generalization of the discrete one, where rather than counting node and edge labelling instances, we count statistics of some representatives of them. We encounter how the proposed vectorial representations of graphs suffer from high dimensionality and correlation among components and we face these problems by feature selection algorithms. We also explore how the diversity of different embedding representations can be exploited in order to boost the performance of base classifiers in a multiple classifier systems framework. An extensive experimental evaluation finally shows how the methodology we propose can be efficiently computed and compete with other graph matching and embedding methodologies.

Abstract: Image segmentations aims to partition an image into a set of non-overlapped regions, called segments. Despite the simplicity of the definition, image segmentation raises as a very complex problem in all its stages. The definition of segment is still unclear. When asking to a human to perform a segmentation, this person segments at different levels of abstraction. Some segments might be a single, well-defined texture whereas some others correspond with an object in the scene which might including multiple textures and colors. For this reason, segmentation is divided in bottom-up segmentation and top-down segmentation. Bottom up-segmentation is problem independent, that is, focused on general properties of the images such as textures or illumination. Top-down segmentation is a problem-dependent approach which looks for specific entities in the scene, such as known objects. This work is focused on bottom-up segmentation. Beginning from the analysis of the lacks of current methods, we propose an approach called RAD. Our approach overcomes the main shortcomings of those methods which use the physics of the light to perform the segmentation. RAD is a topological approach which describes a single-material reflectance. Afterwards, we cope with one of the main problems in image segmentation: non supervised adaptability to image content. To yield a non-supervised method, we use a model of saliency yet presented in this thesis. It computes the saliency of the chromatic transitions of an image by means of a statistical analysis of the images derivatives. This method of saliency is used to build our final approach of segmentation: spRAD. This method is a non-supervised segmentation approach. Our saliency approach has been validated with a psychophysical experiment as well as computationally, overcoming a state-of-the-art saliency method. spRAD also outperforms state-of-the-art segmentation techniques as results obtained with a widely-used segmentation dataset show

Abstract: Humansmove their eyes in order to learn visual representations of the world. These eye movements depend on distinct factors, either by the scene that we perceive or by our own decisions. To select what is relevant to attend is part of our survival mechanisms and the way we build reality, as we constantly react both consciously and unconsciously to all the stimuli that is projected into our eyes. In this thesis we try to explain (1) how we move our eyes, (2) how to build machines that understand visual information and deploy eyemovements, and (3) how to make these machines understand tasks in order to decide for eye movements.
(1) We provided the analysis of eye movement behavior elicited by low-level feature distinctiveness with a dataset of 230 synthetically-generated image patterns. A total of 15 types of stimuli has been generated (e.g. orientation, brightness, color, size, etc.), with 7 feature contrasts for each feature category. Eye-tracking data was collected from 34 participants during the viewing of the dataset, using Free-Viewing and Visual Search task instructions. Results showed that saliency is predominantly and distinctively influenced by: 1. feature type, 2. feature contrast, 3. Temporality of fixations, 4. task difficulty and 5. center bias. From such dataset (SID4VAM), we have computed a benchmark of saliency models by testing performance using psychophysical patterns. Model performance has been evaluated considering model inspiration and consistency with human psychophysics. Our study reveals that state-of-the-art Deep Learning saliency models do not performwell with synthetic pattern images, instead, modelswith Spectral/Fourier inspiration outperform others in saliency metrics and are more consistent with human psychophysical experimentation.
(2) Computations in the primary visual cortex (area V1 or striate cortex) have long been hypothesized to be responsible, among several visual processing mechanisms, of bottom-up visual attention (also named saliency). In order to validate this hypothesis, images from eye tracking datasets have been processed with a biologically plausible model of V1 (named Neurodynamic SaliencyWaveletModel or NSWAM). Following Li’s neurodynamic model, we define V1’s lateral connections with a network of firing rate neurons, sensitive to visual features such as brightness, color, orientation and scale. Early subcortical processes (i.e. retinal and thalamic) are functionally simulated. The resulting saliency maps are generated from the model output, representing the neuronal activity of V1 projections towards brain areas involved in eye movement control. We want to pinpoint that our unified computational architecture is able to reproduce several visual processes (i.e. brightness, chromatic induction and visual discomfort) without applying any type of training or optimization and keeping the same parametrization. The model has been extended (NSWAM-CM) with an implementation of the cortical magnification function to define the retinotopical projections towards V1, processing neuronal activity for each distinct view during scene observation. Novel computational definitions of top-down inhibition (in terms of inhibition of return and selection mechanisms), are also proposed to predict attention in Free-Viewing and Visual Search conditions. Results show that our model outperforms other biologically-inpired models of saliency prediction as well as to predict visual saccade sequences, specifically for nature and synthetic images. We also show how temporal and spatial characteristics of inhibition of return can improve prediction of saccades, as well as how distinct search strategies (in terms of feature-selective or category-specific inhibition) predict attention at distinct image contexts.
(3) Although previous scanpath models have been able to efficiently predict saccades during Free-Viewing, it is well known that stimulus and task instructions can strongly affect eye movement patterns. In particular, task priming has been shown to be crucial to the deployment of eye movements, involving interactions between brain areas related to goal-directed behavior, working and long-termmemory in combination with stimulus-driven eyemovement neuronal correlates. In our latest study we proposed an extension of the Selective Tuning Attentive Reference Fixation ControllerModel based on task demands (STAR-FCT), describing novel computational definitions of Long-TermMemory, Visual Task Executive and Task Working Memory. With these modules we are able to use textual instructions in order to guide the model to attend to specific categories of objects and/or places in the scene. We have designed our memorymodel by processing a visual hierarchy of low- and high-level features. The relationship between the executive task instructions and the memory representations has been specified using a tree of semantic similarities between the learned features and the object category labels. Results reveal that by using this model, the resulting object localizationmaps and predicted saccades have a higher probability to fall inside the salient regions depending on the distinct task instructions compared to saliency.

Abstract: Image generation is arguably one of the most attractive, compelling, and challenging tasks in computer vision. Among the methods which perform image generation, generative adversarial networks (GANs) play a key role. The most common image generation models based on GANs can be divided into two main approaches. The first one, called simply image generation takes random noise as an input and synthesizes an image which follows the same distribution as the images in the training set. The second class, which is called image-to-image translation, aims to map an image from a source domain to one that is indistinguishable from those in the target domain. Image-to-image translation methods can further be divided into paired and unpaired image-to-image translation based on whether they require paired data or not. In this thesis, we aim to address some challenges of both image generation and image-to-image generation.GANs highly rely upon having access to vast quantities of data, and fail to generate realistic images from random noise when applied to domains with few images. To address this problem, we aim to transfer knowledge from a model trained on a large dataset (source domain) to the one learned on limited data (target domain). We find that both GANs andconditional GANs can benefit from models trained on large datasets. Our experiments show that transferring the discriminator is more important than the generator. Using both the generator and discriminator results in the best performance. We found, however, that this method suffers from overfitting, since we update all parameters to adapt to the target data. We propose a novel architecture, which is tailored to address knowledge transfer to very small target domains. Our approach effectively exploreswhich part of the latent space is more related to the target domain. Additionally, the proposed method is able to transfer knowledge from multiple pretrained GANs. Although image-to-image translation has achieved outstanding performance, it still facesseveral problems. First, for translation between complex domains (such as translations between different modalities) image-to-image translation methods require paired data. We show that when only some of the pairwise translations have been seen (i.e. during training), we can infer the remaining unseen translations (where training pairs are not available). We propose a new approach where we align multiple encoders and decoders in such a way that the desired translation can be obtained by simply cascadingthe source encoder and the target decoder, even when they have not interacted during the training stage (i.e. unseen). Second, we address the issue of bias in image-to-image translation. Biased datasets unavoidably contain undesired changes, which are dueto the fact that the target dataset has a particular underlying visual distribution. We use carefully designed semantic constraints to reduce the effects of the bias. The semantic constraint aims to enforce the preservation of desired image properties. Finally, current approaches fail to generate diverse outputs or perform scalable image transfer in a single model. To alleviate this problem, we propose a scalable and diverse image-to-image translation. We employ random noise to control the diversity. The scalabitlity is determined by conditioning the domain label.computer vision, deep learning, imitation learning, adversarial generative networks, image generation, image-to-image translation.