Abstract: Document layout analysis is a known problem to the documents research community and has been vastly explored yielding a multitude of solutions ranging from text mining, and recognition to graph-based representation, visual feature extraction, etc. However, most of the existing works have ignored the crucial fact regarding the scarcity of labeled data. With growing internet connectivity to personal life, an enormous amount of documents had been available in the public domain and thus making data annotation a tedious task. We address this challenge using self-supervision and unlike, the few existing self-supervised document segmentation approaches which use text mining and textual labels, we use a complete vision-based approach in pre-training without any ground-truth label or its derivative. Instead, we generate pseudo-layouts from the document images to pre-train an image encoder to learn the document object representation and localization in a self-supervised framework before fine-tuning it with an object detection model. We show that our pipeline sets a new benchmark in this context and performs at par with the existing methods and the supervised counterparts, if not outperforms. The code is made publicly available at: this https URL

Abstract: Cross-modal retrieval methods have been significantly improved in last years with the use of deep neural networks and large-scale annotated datasets such as ImageNet and Places. However, collecting and annotating such datasets requires a tremendous amount of human effort and, besides, their annotations are limited to discrete sets of popular visual classes that may not be representative of the richer semantics found on large-scale cross-modal retrieval datasets. In this paper, we present a self-supervised cross-modal retrieval framework that leverages as training data the correlations between images and text on the entire set of Wikipedia articles. Our method consists in training a CNN to predict: (1) the semantic context of the article in which an image is more probable to appear as an illustration, and (2) the semantic context of its caption. Our experiments demonstrate that the proposed method is not only capable of learning discriminative visual representations for solving vision tasks like classification, but that the learned representations are better for cross-modal retrieval when compared to supervised pre-training of the network on the ImageNet dataset.

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: Blur detection aims at segmenting the blurred areas of a given image. Recent deep learning-based methods approach this problem by learning an end-to-end mapping between the blurred input and a binary mask representing the localization of its blurred areas. Nevertheless, the effectiveness of such deep models is limited due to the scarcity of datasets annotated in terms of blur segmentation, as blur annotation is labor intensive. In this work, we bypass the need for such annotated datasets for end-to-end learning, and instead rely on object proposals and a model for blur generation in order to produce a dataset of synthetically blurred images. This allows us to perform self-supervised learning over the generated image and ground truth blur mask pairs using CNNs, defining a framework that can be employed in purely self-supervised, weakly supervised or semi-supervised configurations. Interestingly, experimental results of such setups over the largest blur segmentation datasets available show that this approach achieves state of the art results in blur segmentation, even without ever observing any real blurred image.

Abstract: Recent progress in the field of human action recognition points towards the use of Spatio-TemporalInterestPoints (STIPs) for local descriptor-based recognition strategies. In this paper, we present a novel approach for robust and selective STIP detection, by applying surround suppression combined with local and temporal constraints. This new method is significantly different from existing STIP detection techniques and improves the performance by detecting more repeatable, stable and distinctive STIPs for human actors, while suppressing unwanted background STIPs. For action representation we use a bag-of-video words (BoV) model of local N-jet features to build a vocabulary of visual-words. To this end, we introduce a novel vocabulary building strategy by combining spatial pyramid and vocabulary compression techniques, resulting in improved performance and efficiency. Action class specific Support Vector Machine (SVM) classifiers are trained for categorization of human actions. A comprehensive set of experiments on popular benchmark datasets (KTH and Weizmann), more challenging datasets of complex scenes with background clutter and camera motion (CVC and CMU), movie and YouTube video clips (Hollywood 2 and YouTube), and complex scenes with multiple actors (MSR I and Multi-KTH), validates our approach and show state-of-the-art performance. Due to the unavailability of ground truth action annotation data for the Multi-KTH dataset, we introduce an actor specific spatio-temporal clustering of STIPs to address the problem of automatic action annotation of multiple simultaneous actors. Additionally, we perform cross-data action recognition by training on source datasets (KTH and Weizmann) and testing on completely different and more challenging target datasets (CVC, CMU, MSR I and Multi-KTH). This documents the robustness of our proposed approach in the realistic scenario, using separate training and test datasets, which in general has been a shortcoming in the performance evaluation of human action recognition techniques.

Abstract: Dimensionality reduction is key to alleviate machine learning artifacts in clinical applications with Small Sample Size (SSS) unbalanced datasets. Existing methods rely on either the probabilistic distribution of training data or the discriminant power of the reduced space, disregarding the impact of repeatability and uncertainty in features.In the present study is proposed the use of reproducibility of radiomics features to select features with high inter-class correlation coefficient (ICC). The reproducibility includes the variability introduced in the image acquisition, like medical scans acquisition parameters and convolution kernels, that affects intensity-based features and tumor annotations made by physicians, that influences morphological descriptors of the lesion.For the reproducibility of radiomics features three studies were conducted on cases collected at Vall Hebron Oncology Institute (VHIO) on responders to oncology treatment. The studies focused on the variability due to the convolution kernel, image acquisition parameters, and the inter-observer lesion identification. The features selected were those features with a ICC higher than 0.7 in the three studies.The selected features based on reproducibility were evaluated for lesion malignancy classification using a different database. Results show better performance compared to several state-of-the-art methods including Principal Component Analysis (PCA), Kernel Discriminant Analysis via QR decomposition (KDAQR), LASSO, and an own built Convolutional Neural Network.