Abstract: Endoscopy is the process in which a camera is introduced inside a human.
Given that endoscopy provides realistic images (in contrast to other modalities) and allows non-invase minimal intervention procedures (which can aid in diagnosis and surgical interventions), its use has spreaded during last decades.
In this project we will focus on bronchoscopic procedures, during which the camera is introduced through the trachea in order to have a diagnostic of the patient. The diagnostic interventions are focused on: degree of stenosis (reduction in tracheal area), prosthesis or early diagnosis of tumors. In the first case, assessment of the luminal area and the calculation of the diameters of the tracheal rings are required. A main limitation is that all the process is done by hand,
which means that the doctor takes all the measurements and decisions just by looking at the screen. As far as we know there is no computational framework for helping the doctors in the diagnosis.
This project will consist of analysing bronchoscopic videos in order to extract useful information for the diagnostic of the degree of stenosis. In particular we will focus on segmentation of the tracheal rings. As a result of this project several strategies (for detecting tracheal rings) had been implemented in order to compare their performance.

Abstract: Unsupervised representation learning aims at finding methods that learn representations from data without annotation-based signals. Abstaining from annotations not only leads to economic benefits but may – and to some extent already does – result in advantages regarding the representation’s structure, robustness, and generalizability to different tasks. In the long run, unsupervised methods are expected to surpass their supervised counterparts due to the reduction of human intervention and the inherently more general setup that does not bias the optimization towards an objective originating from specific annotation-based signals. While major advantages of unsupervised representation learning have been recently observed in natural language processing, supervised methods still dominate in vision domains for most tasks. In this dissertation, we contribute to the field of unsupervised (visual) representation learning from three perspectives: (i) Learning representations: We design unsupervised, backpropagation-free Convolutional Self-Organizing Neural Networks (CSNNs) that utilize self-organization- and Hebbian-based learning rules to learn convolutional kernels and masks to achieve deeper backpropagation-free models. Thereby, we observe that backpropagation-based and -free methods can suffer from an objective function mismatch between the unsupervised pretext task and the target task. This mismatch can lead to performance decreases for the target task. (ii) Evaluating representations: We build upon the widely used (non-)linear evaluation protocol to define pretext- and target-objective-independent metrics for measuring the objective function mismatch. With these metrics, we evaluate various pretext and target tasks and disclose dependencies of the objective function mismatch concerning different parts of the training and model setup. (iii) Transferring representations: We contribute CARLANE, the first 3-way sim-to-real domain adaptation benchmark for 2D lane detection. We adopt several well-known unsupervised domain adaptation methods as baselines and propose a method based on prototypical cross-domain self-supervised learning. Finally, we focus on pixel-based unsupervised domain adaptation and contribute a content-consistent unpaired image-to-image translation method that utilizes masks, global and local discriminators, and similarity sampling to mitigate content inconsistencies, as well as feature-attentive denormalization to fuse content-based statistics into the generator stream. In addition, we propose the cKVD metric to incorporate class-specific content inconsistencies into perceptual metrics for measuring translation quality.

Abstract: The recognition of handwritten music scores still remains an open problem. The existing approaches can only deal with very simple handwritten scores mainly because of the variability in the handwriting style and the variability in the composition of groups of music notes (i.e. compound music notes). In this work we focus on this second problem and propose a method based on perceptual grouping for the recognition of compound music notes. Our method has been tested using several handwritten music scores of the CVC-MUSCIMA database and compared with a commercial Optical Music Recognition (OMR) software. Given that our method is learning-free, the obtained results are promising.

Abstract: It is very common to nd dierent versions of the same music work in archives of Opera Theaters. These dierences correspond to modications and annotations from the musicians. From the musicologist point of view, these variations are very interesting and deserve study.
This paper explores the alignment of music scores as a tool for automatically detecting the passages that contain such dierences. Given the diculties in the recognition of handwritten music scores, our goal is to align the music scores and at the same time, avoid the recognition of music elements as much as possible. After removing the sta lines, braces and ties, the bar lines are detected. Then, the bar units are described as a whole using the Blurred Shape Model. The bar units alignment is performed by using Dynamic Time Warping. The analysis of the alignment path is used to detect the variations in the music scores. The method has been evaluated on a subset of the CVC-MUSCIMA dataset, showing encouraging results.

Abstract: Visual lifelogging consists of acquiring images that capture the daily experiences of the user by wearing a camera over a long period of time. The pictures taken offer considerable potential for knowledge mining concerning how people live their lives, hence, they open up new opportunities for many potential applications in fields including healthcare, security, leisure and
the quantified self. However, automatically building a story from a huge collection of unstructured egocentric data presents major challenges. This paper provides a thorough review of advances made so far in egocentric data analysis, and in view of the current state of the art, indicates new lines of research to move us towards storytelling from visual lifelogging.

Abstract: Following the increasingly popular trend of social interaction analysis in egocentric vision, this article presents a comprehensive pipeline for automatic social pattern characterization of a wearable photo-camera user. The proposed framework relies merely on the visual analysis of egocentric photo-streams and consists of three major steps. The first step is to detect social interactions of the user where the impact of several social signals on the task is explored. The detected social events are inspected in the second step for categorization into different social meetings. These two steps act at event-level where each potential social event is modeled as a multi-dimensional time-series, whose dimensions correspond to a set of relevant features for each task; finally, LSTM is employed to classify the time-series. The last step of the framework is to characterize social patterns of the user. Our goal is to quantify the duration, the diversity and the frequency of the user social relations in various social situations. This goal is achieved by the discovery of recurrences of the same people across the whole set of social events related to the user. Experimental evaluation over EgoSocialStyle – the proposed dataset in this work, and EGO-GROUP demonstrates promising results on the task of social pattern characterization from egocentric photo-streams.

Abstract: Artificial intelligence is innovating the fashion industry by proposing new applications and solutions to the problems encountered by researchers and engineers working in the industry. In this thesis, we address three of these problems. In the first part of the thesis, we tackle the problem of multi-label image classification which is very related to fashion attribute recognition. In the second part of the thesis, we address two problems that are specific to fashion. Firstly, we address the problem of main product detection which is the task of associating correct image parts (e.g. bounding boxes) with the fashion product being sold. Secondly, we address the problem of color naming for multicolored fashion items. The task of multi-label image classification consists in assigning various concepts such as objects or attributes to images. Usually, there are dependencies that can be learned between the concepts to capture label correlations (chair and table classes are more likely to co-exist than chair and giraffe).
If we treat the multi-label image classification problem as an orderless set prediction problem, we can exploit recurrent neural networks (RNN) to capture label correlations. However, RNNs are trained to predict ordered sequences of tokens, so if the order of the predicted sequence is different than the order of the ground truth sequence, there will be penalization although the predictions are correct. Therefore, in the first part of the thesis, we propose an orderless loss function which will order the labels in the ground truth sequence dynamically in a way that the minimum loss is achieved. This results in a significant improvement of RNN models on multi-label image classification over the previous methods.
However, RNNs suffer from long term dependencies when the cardinality of set grows bigger. The decoding process might stop early if the current hidden state cannot find any object and outputs the termination token. This would cause the remaining classes not to be predicted and lower recall metric. Transformers can be used to avoid the long term dependency problem exploiting their selfattention modules that process sequential data simultaneously. Consequently, we propose a novel transformer model for multi-label image classification which surpasses the state-of-the-art results by a large margin.
In the second part of thesis, we focus on two fashion-specific problems. Main product detection is the task of associating image parts with the fashion product that is being sold, generally using associated textual metadata (product title or description). Normally, in fashion e-commerces, products are represented by multiple images where a person wears the product along with other fashion items. If all the fashion items in the images are marked with bounding boxes, we can use the textual metadata to decide which item is the main product. The initial work treated each of these images independently, discarding the fact that they all belong to the same product. In this thesis, we represent the bounding boxes from all the images as nodes in a fully connected graph. This allows the algorithm to learn relations between the nodes during training and take the entire context into account for the final decision. Our algorithm results in a significant improvement of the state-ofthe-art.
Moreover, we address the problem of color naming for multicolored fashion items, which is a challenging task due to the external factors such as illumination changes or objects that act as clutter. In the context of multi-label classification, the vaguely defined lines between the classes in the color space cause ambiguity. For example, a shade of blue which is very close to green might cause the model to incorrectly predict the color blue and green at the same time. Based on this, models trained for color naming are expected to recognize the colors and their quantities in both single colored and multicolored fashion items. Therefore, in this thesis, we propose a novel architecture with an additional head that explicitly estimates the number of colors in fashion items. This removes the ambiguity problem and results in better color naming performance.