Abstract: Deep learning has recently been enjoying an increasing popularity due to its success in solving challenging tasks. In particular, deep learning has proven to be effective in a large variety of computer vision tasks, such as image classification, object recognition and image parsing. Contrary to previous research, which required engineered feature representations, designed by experts, in order to succeed, deep learning attempts to learn representation hierarchies automatically from data. More recently, the trend has been to go deeper with representation hierarchies.
Learning (very) deep representation hierarchies is a challenging task, which
involves the optimization of highly non-convex functions. Therefore, the search
for algorithms to ease the learning of (very) deep representation hierarchies from data is extensive and ongoing.
In this thesis, we tackle the challenging problem of easing the learning of (very) deep representation hierarchies. We present a hyper-parameter free, off-the-shelf, simple and fast unsupervised algorithm to discover hidden structure from the input data by enforcing a very strong form of sparsity. We study the applicability and potential of the algorithm to learn representations of varying depth in a handful of applications and domains, highlighting the ability of the algorithm to provide discriminative feature representations that are able to achieve top performance.
Yet, while emphasizing the great value of unsupervised learning methods when
labeled data is scarce, the recent industrial success of deep learning has revolved around supervised learning. Supervised learning is currently the focus of many recent research advances, which have shown to excel at many computer vision tasks. Top performing systems often involve very large and deep models, which are not well suited for applications with time or memory limitations. More in line with the current trends, we engage in making top performing models more efficient, by designing very deep and thin models. Since training such very deep models still appears to be a challenging task, we introduce a novel algorithm that guides the training of very thin and deep models by hinting their intermediate representations.
Very deep and thin models trained by the proposed algorithm end up extracting feature representations that are comparable or even better performing
than the ones extracted by large state-of-the-art models, while compellingly
reducing the time and memory consumption of the model.

Abstract: In the early twenty first century Given Imaging Ltd. presented wireless capsule endoscopy (WCE) as a new technological breakthrough that allowed the visualization of
the intestine by using a small, swallowed camera. This small size device was received
with a high enthusiasm within the medical community, and until now, it is still one
of the medical devices with the highest use growth rate. WCE can be used as a novel
diagnostic tool that presents several clinical advantages, since it is non-invasive and
at the same time it provides, for the first time, a full picture of the small bowel morphology, contents and dynamics. Since its appearance, the WCE has been used to
detect several intestinal dysfunctions such as: polyps, ulcers and bleeding. However,
the visual analysis of WCE videos presents an important drawback: the long time
required by the physicians for proper video visualization. In this sense and regarding
to this limitation, the development of computer aided systems is required for the extensive use of WCE in the medical community.
The work presented in this thesis is a set of contributions for the automatic image
analysis and computer-aided diagnosis of intestinal motility disorders using WCE.
Until now, the diagnosis of small bowel motility dysfunctions was basically performed
by invasive techniques such as the manometry test, which can only be conducted at
some referral centers around the world owing to the complexity of the procedure and
the medial expertise required in the interpretation of the results.
Our contributions are divided in three main blocks:
1. Image analysis by computer vision techniques to detect events in the endoluminal WCE scene. Several methods have been proposed to detect visual events
such as: intestinal contractions, intestinal content, tunnel and wrinkles;
2. Machine learning techniques for the analysis and the manipulation of the data
from WCE. These methods have been proposed in order to overcome the problems that the analysis of WCE presents such as: video acquisition cost, unlabeled data and large number of data;
3. Two different systems for the computer-aided diagnosis of intestinal motility
disorders using WCE. The first system presents a fully automatic method that
aids at discriminating healthy subjects from patients with severe intestinal motor disorders like pseudo-obstruction or food intolerance. The second system presents another automatic method that models healthy subjects and discriminate them from mild intestinal motility patients.

Abstract: Pedestrian detection is of paramount interest for many applications, e.g. Advanced Driver Assistance Systems, Intelligent Video Surveillance and Multimedia systems. Most promising pedestrian detectors rely on appearance-based classifiers trained with annotated data. However, the required annotation step represents an intensive and subjective task for humans, what makes worth to minimize their intervention in this process by using computational tools like realistic virtual worlds. The reason to use these kind of tools relies in the fact that they allow the automatic generation of precise and rich annotations of visual information. Nevertheless, the use of this kind of data comes with the following question: can a pedestrian appearance model learnt with virtual-world data work successfully for pedestrian detection in real-world scenarios?. To answer this question, we conduct different experiments that suggest a positive answer. However, the pedestrian classifiers trained with virtual-world data can suffer the so called dataset shift problem as real-world based classifiers does. Accordingly, we have designed different domain adaptation techniques to face this problem, all of them integrated in a same framework (V-AYLA). We have explored different methods to train a domain adapted pedestrian classifiers by collecting a few pedestrian samples from the target domain (real world) and combining them with many samples of the source domain (virtual world). The extensive experiments we present show that pedestrian detectors developed within the V-AYLA framework do achieve domain adaptation. Ideally, we would like to adapt our system without any human intervention. Therefore, as a first proof of concept we also propose an unsupervised domain adaptation technique that avoids human intervention during the adaptation process. To the best of our knowledge, this Thesis work is the first demonstrating adaptation of virtual and real worlds for developing an object detector. Last but not least, we also assessed a different strategy to avoid the dataset shift that consists in collecting real-world samples and retrain with them in such a way that no bounding boxes of real-world pedestrians have to be provided. We show that the generated classifier is competitive with respect to the counterpart trained with samples collected by manually annotating pedestrian bounding boxes. The results presented on this Thesis not only end with a proposal for adapting a virtual-world pedestrian detector to the real world, but also it goes further by pointing out a new methodology that would allow the system to adapt to different situations, which we hope will provide the foundations for future research in this unexplored area.

Abstract: Action recognition is a very challenging and important problem in computer vi­sion. Researchers working on this field aspire to provide computers with the abil­ ity to visually perceive human actions – that is, to observe, interpret, and under­ stand human-related events that occur in the physical environment merely from visual data. The applications of this technology are numerous: human-machine interaction, e-health, monitoring/surveillance, and content-based video retrieval, among others. Hand-crafted methods dominated the field until the apparition of the first successful deep learning-based action recognition works. Although ear­ lier deep-based methods underperformed with respect to hand-crafted approaches, these slowly but steadily improved to become state-of-the-art, eventually achieving better results than hand-crafted ones. Still, hand-crafted approaches can be advan­ tageous in certain scenarios, specially when not enough data is available to train very large deep models or simply to be combined with deep-based methods to fur­ ther boost the performance. Hence, showing how hand-crafted features can provide extra knowledge the deep networks are notable to easily learn about human actions.
This Thesis concurs in time with this change of paradigm and, hence, reflects it into two distinguished parts. In the first part, we focus on improving current suc­ cessful hand-crafted approaches for action recognition and we do so from three dif­ ferent perspectives. Using the dense trajectories framework as a backbone: first, we explore the use of multi-modal and multi-view input
data to enrich the trajectory de­ scriptors. Second, we focus on the classification part of action recognition pipelines and propose an ensemble learning approach, where each classifier leams from a dif­ferent set of local spatiotemporal features to then combine their outputs following an strategy based on the Dempster-Shaffer Theory. And third, we propose a novel hand-crafted feature extraction method that constructs a rnid-level feature descrip­ tion to better modellong-term spatiotemporal dynarnics within action videos. Moving to the second part of the Thesis, we start with a comprehensive study of the current deep-learning based action recognition methods. We review both fun­ damental and cutting edge methodologies reported during the last few years and introduce a taxonomy of deep-leaming methods dedicated to action recognition. In particular, we analyze and discuss how these handle
the temporal dimension of data. Last but not least, we propose a residual recurrent network for action recogni­ tion that naturally integrates all our previous findings in a powerful and prornising framework.

Abstract: Wireless Capsule Endoscopy (WCE) is a technique for inner-visualization of the entire small intestine and, thus, offers an interesting perspective on intestinal motility. The two major drawbacks of this technique are: 1) huge amount of data acquired by WCE makes the motility analysis tedious and 2) since the capsule is the first tool that offers complete inner-visualization of the small intestine,the exact importance of the observed events is still an open issue. Therefore, in this thesis, a novel computer-aided system for intestinal motility analysis is presented. The goal of the system is to provide an easily-comprehensible visual description of motility-related intestinal events to a physician. In order to do so, several tools based either on computer vision concepts or on machine learning techniques are presented. A method for transforming 3D video signal to a holistic image of intestinal motility, called motility bar, is proposed. The method calculates the optimal mapping from video into image from the intestinal motility point of view.
To characterize intestinal motility, methods for automatic extraction of motility information from WCE are presented. Two of them are based on the motility bar and two of them are based on frame-per-frame analysis. In particular, four algorithms dealing with the problems of intestinal contraction detection, lumen size estimation, intestinal content characterization and wrinkle frame detection are proposed and validated. The results of the algorithms are converted into sequential features using an online statistical test. This test is designed to work with multivariate data streams. To this end, we propose a novel formulation of concentration inequality that is introduced into a robust adaptive windowing algorithm for multivariate data streams. The algorithm is used to obtain robust representation of segments with constant intestinal motility activity. The obtained sequential features are shown to be discriminative in the problem of abnormal motility characterization.
Finally, we tackle the problem of efficient labeling. To this end, we incorporate active learning concepts to the problems present in WCE data and propose two approaches. The first one is based the concepts of sequential learning and the second one adapts the partition-based active learning to an error-free labeling scheme. All these steps are sufficient to provide an extensive visual description of intestinal motility that can be used by an expert as decision support system.