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 registration has been for many years the gold standard method to bring two images into correspondence. It has been used extensively in the eld of medical imaging in order to put images of dierent patients into a common overlapping spatial position. However, medical image registration is a slow, iterative optimization process, where many variables and prone to fall into the pit traps local minima.
A coordinate system parameterizing the interior of organs is a powerful tool for a systematic localization of injured tissue. If the same coordinate values are assigned to specic anatomical sites, parameterizations ensure integration of data across different medical image modalities. Harmonic mappings have been used to produce parametric meshes over the surface of anatomical shapes, given their ability to set values at specic locations through boundary conditions. However, most of the existing implementations in medical imaging restrict to either anatomical surfaces, or the depth coordinate with boundary conditions is given at discrete sites of limited geometric diversity.
The medial surface of the shape can be used to provide a continuous basis for the denition of a depth coordinate. However, given that dierent methods for generation of medial surfaces generate dierent manifolds, not all of them are equally suited to be the basis of radial coordinate for a parameterization. It would be desirable that the medial surface will be smooth, and robust to surface shape noise, with low number of spurious branches or surfaces.
In this thesis we present methods for computation of smooth medial manifolds and apply them to the generation of for anatomical volumetric parameterization that extends current harmonic parameterizations to the interior anatomy using information provided by the volume medial surface. This reference system sets a solid base for creating anatomical models of the anatomical shapes, and allows comparing several patients in a common framework of reference.

Abstract: Image values are the result of a combination of visual information coming from multiple sources. Recovering information from the multiple factors thatproduced an image seems a hard and ill-posed problem. However, it is important to observe that humans develop the ability to interpret images and recognize and isolate specific physical properties of the scene.
Images describing a single physical characteristic of an scene are called intrinsic images. These images would benefit most computer vision tasks which are often affected by the multiple complex effects that are usually found in natural images (e.g. cast shadows, specularities, interreflections...).
In this thesis we analyze the problem of intrinsic image estimation from different perspectives, including the theoretical formulation of the problem, the visual cues that can be used to estimate the intrinsic components and the evaluation mechanisms of the problem.

Abstract: In autonomous driving, artificial intelligence (AI) processes the traffic environment to drive the vehicle to a desired destination. Currently, there are different paradigms that address the development of AI-enabled drivers. On the one hand, we find modular pipelines, which divide the driving task into sub-tasks such as perception, maneuver planning, and control. On the other hand, we find end-to-end driving approaches that attempt to learn the direct mapping of raw data from input sensors to vehicle control signals. The latter are relatively less studied but are gaining popularity as they are less demanding in terms of data labeling. Therefore, in this thesis, our goal is to investigate end-to-end autonomous driving.
We propose to evaluate three approaches to tackle the challenge of end-to-end
autonomous driving. First, we focus on the input, considering adding depth information as complementary to RGB data, in order to mimic the human being’s
ability to estimate the distance to obstacles. Notice that, in the real world, these depth maps can be obtained either from a LiDAR sensor, or a trained monocular
depth estimation module, where human labeling is not needed. Then, based on
the intuition that the latent space of end-to-end driving models encodes relevant
information for driving, we use it as prior knowledge for training an affordancebased driving model. In this case, the trained affordance-based model can achieve good performance while requiring less human-labeled data, and it can provide interpretability regarding driving actions. Finally, we present a new pure vision-based end-to-end driving model termed CIL++, which is trained by imitation learning.
CIL++ leverages modern best practices, such as a large horizontal field of view and
a self-attention mechanism, which are contributing to the agent’s understanding of
the driving scene and bringing a better imitation of human drivers. Using training
data without any human labeling, our model yields almost expert performance in
the CARLA NoCrash benchmark and could rival SOTA models that require large amounts of human-labeled data.

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: In this dissertation, we explore different ways to perform human action recognition in video clips. We focus on data efficiency, proposing new approaches that alleviate the need for laborious and time-consuming manual data annotation. In the first part of this dissertation, we start by analyzing previous state-of-the-art models, comparing their differences and similarities in order to pinpoint where their real strengths come from. Leveraging this information, we then proceed to boost the classification accuracy of shallow models to levels that rival deep neural networks. We introduce hybrid video classification architectures based on carefully designed unsupervised representations of handcrafted spatiotemporal features classified by supervised deep networks. We show in our experiments that our hybrid model combine the best of both worlds: it is data efficient (trained on 150 to 10,000 short clips) and yet improved significantly on the state of the art, including deep models trained on millions of manually labeled images and videos. In the second part of this research, we investigate the generation of synthetic training data for action recognition, as it has recently shown promising results for a variety of other computer vision tasks. We propose an interpretable parametric generative model of human action videos that relies on procedural generation and other computer graphics techniques of modern game engines. We generate a diverse, realistic, and physically plausible dataset of human action videos, called PHAV for “Procedural Human Action Videos”. It contains a total of 39,982 videos, with more than 1,000 examples for each action of 35 categories. Our approach is not limited to existing motion capture sequences, and we procedurally define 14 synthetic actions. We then introduce deep multi-task representation learning architectures to mix synthetic and real videos, even if the action categories differ. Our experiments on the UCF-101 and HMDB-51 benchmarks suggest that combining our large set of synthetic videos with small real-world datasets can boost recognition performance, outperforming fine-tuning state-of-the-art unsupervised generative models of videos.

Abstract: In this thesis we explore information Extraction from totally or partially handwritten documents. Basically we are dealing with two different application scenarios. The first scenario are modern highly structured documents like forms. In this kind of documents, the semantic information is encoded in different fields with a pre-defined location in the document, therefore, information extraction becomes roughly equivalent to transcription. The second application scenario are loosely structured totally handwritten documents, besides transcribing them, we need to assign a semantic label, from a set of known values to the handwritten words.
In both scenarios, transcription is an important part of the information extraction. For that reason in this thesis we present two methods based on Neural Networks, to transcribe handwritten text.In order to tackle the challenge of loosely structured documents, we have produced a benchmark, consisting of a dataset, a defined set of tasks and a metric, that was presented to the community as an international competition. Also, we propose different models based on Convolutional and Recurrent neural networks that are able to transcribe and assign different semantic labels to each handwritten words, that is, able to perform Information Extraction.

Abstract: In this thesis, several methods for the automatic analysis of Intravascular Ultrasound
(IVUS) sequences are presented, aimed at assisting physicians in the diagnosis, the assessment of the intervention and the monitoring of the patients with coronary disease.
The basis for the developed frameworks are machine learning, pattern recognition and
image processing techniques.
First, a novel approach for the automatic detection of vascular bifurcations in
IVUS is presented. The task is addressed as a binary classication problem (identifying bifurcation and non-bifurcation angular sectors in the sequence images). The
multiscale stacked sequential learning algorithm is applied, to take into account the
spatial and temporal context in IVUS sequences, and the results are rened using
a-priori information about branching dimensions and geometry. The achieved performance is comparable to intra- and inter-observer variability.
Then, we propose a novel method for the automatic non-rigid alignment of IVUS
sequences of the same patient, acquired at dierent moments (before and after percutaneous coronary intervention, or at baseline and follow-up examinations). The
method is based on the description of the morphological content of the vessel, obtained by extracting temporal morphological proles from the IVUS acquisitions, by
means of methods for segmentation, characterization and detection in IVUS. A technique for non-rigid sequence alignment – the Dynamic Time Warping algorithm -
is applied to the proles and adapted to the specic clinical problem. Two dierent robust strategies are proposed to address the partial overlapping between frames
of corresponding sequences, and a regularization term is introduced to compensate
for possible errors in the prole extraction. The benets of the proposed strategy
are demonstrated by extensive validation on synthetic and in-vivo data. The results
show the interest of the proposed non-linear alignment and the clinical value of the
method.
Finally, a novel automatic approach for the extraction of the luminal border in
IVUS images is presented. The method applies the multiscale stacked sequential
learning algorithm and extends it to 2-D+T, in a rst classication phase (the identi-
cation of lumen and non-lumen regions of the images), while an active contour model
is used in a second phase, to identify the lumen contour. The method is extended
to the longitudinal dimension of the sequences and it is validated on a challenging
data-set.