Abstract: The mass irruption of Deep Convolutional Neural Networks (CNNs) in computer vision since 2012 led to a dominance of the image understanding paradigm consisting in an end-to-end fully supervised learning workflow over large-scale annotated datasets. This approach proved to be extremely useful at solving a myriad of classic and new computer vision tasks with unprecedented performance —often, surpassing that of humans—, at the expense of vast amounts of human-labeled data, extensive computational resources and the disposal of all of our prior knowledge on the task at hand. Even though simple transfer learning methods, such as fine-tuning, have achieved remarkable impact, their success when the amount of labeled data in the target domain is small is limited. Furthermore, the non-static nature of data generation sources will often derive in data distribution shifts that degrade the performance of deployed models. As a consequence, there is a growing demand for methods that can exploit elements of prior knowledge and sources of information other than the manually generated ground truth annotations of the images during the network training process, so that they can adapt to new domains that constitute, if not a small data regime, at least a small labeled data regime. This thesis targets such few or no labeled data scenario in three distinct image-to-image mapping learning problems. It contributes with various approaches that leverage our previous knowledge of different elements of the image formation process: We first present a data-efficient framework for both defocus and motion blur detection, based on a model able to produce realistic synthetic local degradations. The framework comprises a self-supervised, a weakly-supervised and a semi-supervised instantiation, depending on the absence or availability and the nature of human annotations, and outperforms fully-supervised counterparts in a variety of settings. Our knowledge on color image formation is then used to gather input and target ground truth image pairs for the RGB to hyperspectral image reconstruction task. We make use of a CNN to tackle this problem, which, for the first time, allows us to exploit spatial context and achieve state-of-the-art results given a limited hyperspectral image set. In our last contribution to the subfield of data-efficient image-to-image transformation problems, we present the novel semi-supervised task of zero-pair cross-view semantic segmentation: we consider the case of relocation of the camera in an end-to-end trained and deployed monocular, fixed-view semantic segmentation system often found in industry. Under the assumption that we are allowed to obtain an additional set of synchronized but unlabeled image pairs of new scenes from both original and new camera poses, we present ZPCVNet, a model and training procedure that enables the production of dense semantic predictions in either source or target views at inference time. The lack of existing suitable public datasets to develop this approach led us to the creation of MVMO, a large-scale Multi-View, Multi-Object path-traced dataset with per-view semantic segmentation annotations. We expect MVMO to propel future research in the exciting under-developed fields of cross-view and multi-view semantic segmentation. Last, in a piece of applied research of direct application in the context of process monitoring of an Electric Arc Furnace (EAF) in a steelmaking plant, we also consider the problem of simultaneously estimating the temperature and spectral emissivity of distant hot emissive samples. To that end, we design our own capturing device, which integrates three point spectrometers covering a wide range of the Ultra-Violet, visible, and Infra-Red spectra and is capable of registering the radiance signal incoming from an 8cm diameter spot located up to 20m away. We then define a physically accurate radiative transfer model that comprises the effects of atmospheric absorbance, of the optical system transfer function, and of the sample temperature and spectral emissivity themselves. We solve this inverse problem without the need for annotated data using a probabilistic programming-based Bayesian approach, which yields full posterior distribution estimates of the involved variables that are consistent with laboratory-grade measurements.

Abstract: Optical Flow (OF) is the input of a wide range of decision support systems such as car driver assistance, UAV guiding or medical diagnose. In these real situations, the absence of ground truth forces to assess OF quality using quantities computed from either sequences or the computed optical flow itself. These quantities are generally known as Confidence Measures, CM. Even if we have a proper confidence measure we still need a way to evaluate its ability to discard pixels with an OF prone to have a large error. Current approaches only provide a descriptive evaluation of the CM performance but such approaches are not capable to fairly compare different confidence measures and optical flow algorithms. Thus, it is of prime importance to define a framework and a general road map for the evaluation of optical flow performance.
This thesis provides a framework able to decide which pairs “ optical flow – confidence measure” (OF-CM) are best suited for optical flow error bounding given a confidence level determined by a decision support system. To design this framework we cover the following points:
Descriptive scores. As a first step, we summarize and analyze the sources of inaccuracies in the output of optical flow algorithms. Second, we present several descriptive plots that visually assess CM capabilities for OF error bounding. In addition to the descriptive plots, given a plot representing OF-CM capabilities to bound the error, we provide a numeric score that categorizes the plot according to its decreasing profile, that is, a score assessing CM performance.
Statistical framework. We provide a comparison framework that assesses the best suited OF-CM pair for error bounding that uses a two stage cascade process. First of all we assess the predictive value of the confidence measures by means of a descriptive plot. Then, for a sample of descriptive plots computed over training frames, we obtain a generic curve that will be used for sequences with no ground truth. As a second step, we evaluate the obtained general curve and its capabilities to really reflect the predictive value of a confidence measure using the variability across train frames by means of ANOVA.
The presented framework has shown its potential in the application on clinical decision support systems. In particular, we have analyzed the impact of the different image artifacts such as noise and decay to the output of optical flow in a cardiac diagnose system and we have improved the navigation inside the bronchial tree on bronchoscopy.

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: Despite decades of research in Optical Music Recognition (OMR), the recognition of old handwritten music scores remains a challenge because of the variabilities in the handwriting styles, paper degradation, lack of standard notation, etc. Therefore, the research in OMR systems adapted to the particularities of old manuscripts is crucial to accelerate the conversion of music scores existing in archives into digital libraries, fostering the dissemination and preservation of our music heritage. In this paper we explore the adaptation of sequence-to-sequence models with attention mechanism (used in translation and handwritten text recognition) and the generation of specific synthetic data for recognizing old music scores. The experimental validation demonstrates that our approach is promising, especially when compared with long short-term memory neural networks.

Abstract: Metric learning refers to the problem in machine learning of learning a distance or similarity measurement to compare data. In particular, deep metric learning involves learning a representation, also referred to as embedding, such that in the embedding space data samples can be compared based on the distance, directly providing a similarity measure. This step is necessary to perform several tasks in computer vision. It allows to perform the classification of images, regions or pixels, re-identification, out-of-distribution detection, object tracking in image sequences and any other task that requires computing a similarity score for their solution. This thesis addresses three specific problems that share this common requirement. The first one is person re-identification. Essentially, it is an image retrieval task that aims at finding instances of the same person according to a similarity measure. We first compare in terms of accuracy and efficiency, classical metric learning to basic deep learning based methods for this problem. In this context, we also study network distillation as a strategy to optimize the trade-off between accuracy and speed at inference time. The second problem we contribute to is novelty detection in image classification. It consists in detecting samples of novel classes, i.e. never seen during training. However, standard novelty detection does not provide any information about the novel samples besides they are unknown. Aiming at more informative outputs, we take advantage from the hierarchical taxonomies that are intrinsic to the classes. We propose a metric learning based approach that leverages the hierarchical relationships among classes during training, being able to predict the parent class for a novel sample in such hierarchical taxonomy. Our third contribution is in multi-object tracking and segmentation. This joint task comprises classification, detection, instance segmentation and tracking. Tracking can be formulated as a retrieval problem to be addressed with metric learning approaches. We tackle the existing difficulty in academic research that is the lack of annotated benchmarks for this task. To this matter, we introduce the problem of weakly supervised multi-object tracking and segmentation, facing the challenge of not having available ground truth for instance segmentation. We propose a synergistic training strategy that benefits from the knowledge of the supervised tasks that are being learnt simultaneously.

Abstract: Hand gesture recognition from sequences of depth maps is a challenging computer vision task because of the low inter-class and high intra-class variability, different execution rates of each gesture, and the high articulated nature of human hand. In this paper, a multilevel temporal sampling (MTS) method is first proposed that is based on the motion energy of key-frames of depth sequences. As a result, long, middle, and short sequences are generated that contain the relevant gesture information. The MTS results in increasing the intra-class similarity while raising the inter-class dissimilarities. The weighted depth motion map (WDMM) is then proposed to extract the spatio-temporal information from generated summarized sequences by an accumulated weighted absolute difference of consecutive frames. The histogram of gradient (HOG) and local binary pattern (LBP) are exploited to extract features from WDMM. The obtained results define the current state-of-the-art on three public benchmark datasets of: MSR Gesture 3D, SKIG, and MSR Action 3D, for 3D hand gesture recognition. We also achieve competitive results on NTU action dataset.

Abstract: The Barcelona Historical Marriage Database (BHMD) gathers records of the more than 600,000 marriages celebrated in the Diocese of Barcelona and their taxation registered in Barcelona Cathedral's so-called Marriage Licenses Books for the long period 1451–1905 and the BALL Demographic Database brings together the individual information recorded in the population registers, censuses and fiscal censuses of the main municipalities of the county of Baix Llobregat (Barcelona). In this ongoing collection 263,786 individual observations have been assembled, dating from the period between 1828 and 1965 by December 2020. The two databases started as part of different interdisciplinary research projects at the crossroads of Historical Demography and Computer Vision. Their construction uses artificial intelligence and computer vision methods as Handwriting Recognition to reduce the time of execution. However, its current state still requires some human intervention which explains the implemented crowdsourcing and game sourcing experiences. Moreover, knowledge graph techniques have allowed the application of advanced record linkage to link the same individuals and families across time and space. Moreover, we will discuss the main research lines using both databases developed so far in historical demography.