Knowledge distillation (KD) remains challenging due to the opaque nature of the knowledge transfer process from a Teacher to a Student, making it difficult to address certain issues related to KD. To address this, we proposed UniCAM, a novel gradient-based visual explanation method, which effectively interprets the knowledge learned during KD. Our experimental results demonstrate that with the guidance of the Teacher's knowledge, the Student model becomes more efficient, learning more relevant features while discarding those that are not relevant. We refer to the features learned with the Teacher's guidance as distilled features and the features irrelevant to the task and ignored by the Student as residual features. Distilled features focus on key aspects of the input, such as textures and parts of objects. In contrast, residual features demonstrate more diffused attention, often targeting irrelevant areas, including the backgrounds of the target objects. In addition, we proposed two novel metrics: the feature similarity score (FSS) and the rele-vance score (RS), which quantify the relevance of the dis-tilled knowledge. Experiments on the CIFAR10, ASIRRA, and Plant Disease datasets demonstrate that UniCAM and the two metrics offer valuable insights to explain the KD process.

Glaucoma, a leading cause of irreversible blindness, necessitates early detection for accurate and timely intervention to prevent irreversible vision loss. In this study, we present a novel deep learning framework that leverages the diagnostic value of 3D Optical Coherence Tomography (OCT) imaging for automated glaucoma detection. In this framework, we integrate a pre-trained Vision Transformer on retinal data for rich slice-wise feature extraction and a bidirectional Gated Recurrent Unit for capturing inter-slice spatial dependencies. This dual-component approach enables comprehensive analysis of local nuances and global structural integrity, crucial for accurate glaucoma diagnosis. Experimental results on a large dataset demonstrate the superior performance of the proposed method over state-of-the-art ones, achieving an F1-score of 93.01%, Matthews Correlation Coefficient (MCC) of 69.33%, and AUC of 94.20%. The framework's ability to leverage the valuable information in 3D OCT data holds significant potential for enhancing clinical decision support systems and improving patient outcomes in glaucoma management.

Glaucoma, a leading cause of irreversible blindness worldwide, poses significant diagnostic challenges due to its reliance on subjective evaluation. Recent advances in computer vision and deep learning have demonstrated the potential for automated assessment. This paper provides a comprehensive survey of studies on AI-based glaucoma diagnosis using fundus, optical coherence tomography, and visual field images, with a focus on deep learning-based methods. We searched Web of Science, PubMed, IEEE Xplore, and Google Scholar, applying specific selection criteria to identify relevant studies published from 2017 to 2023. Our analysis provides a structured overview of architectural paradigms, including convolutional neural networks, autoencoders, attention networks, generative adversarial networks, and geometric deep learning models. Additionally, we discuss approaches for extracting informative features, such as structural, statistical, and hybrid techniques. Furthermore, we outline key research challenges and future directions, emphasizing the need for larger, more diverse datasets, strategies for early disease detection, multi-modal data integration, model explainability, and clinical translation. This survey is expected to be useful for Artificial Intelligence (AI) researchers seeking to translate advances into practice and ophthalmologists aiming to improve clinical workflows and diagnosis using the latest AI outcomes.

Nowadays, a significant part of our time is spent sharing multimodal data on social media sites such as Instagram, Facebook and Twitter. The particular way through which users present themselves to social media can provide useful insights into their behaviours, personalities, perspectives, motives and needs. This article proposes to use multimodal data collected from Instagram accounts to predict the five basic prototypical needs described in Glasser's choice theory (i.e.,Survival,Power,Freedom,Belonging, andFun). We automate the identification of the unconsciously perceived needs from Instagram profiles by using both visual and textual contents. The proposed approach aggregates the visual and textual features extracted using deep learning and constructs a homogeneous representation for each profile through the proposedBag-of-Content. Finally, we perform multi-label classification on the fusion of both modalities. We validate our proposal on a large database, consensually annotated by two expert psychologists, with more than 30,000 images, captions and comments. Experiments show promising accuracy and complementary information between visual and textual cues.

Automated emotion recognition in the wild from facial images remains a challenging problem. Although recent advances in deep learning have assumed a significant breakthrough in this topic, strong changes in pose, orientation, and point of view severely harm current approaches. In addition, the acquisition of labeled datasets is costly and the current state-of-the-art deep learning algorithms cannot model all the aforementioned difficulties. In this article, we propose applying a multitask learning loss function to share a common feature representation with other related tasks. Particularly, we show that emotion recognition benefits from jointly learning a model with a detector of facial action units (collective muscle movements). The proposed loss function addresses the problem of learning multiple tasks with heterogeneously labeled data, improving previous multitask approaches. We validate the proposal using three datasets acquired in noncontrolled environments, and an application to predict compound facial emotion expressions.

Automated emotion recognition from facial images is an unsolved problem in computer vision. Although recent methods achieve close to human accuracy in controlled scenarios, the recognition of emotions in the wild remains a challenging problem. Recent advances in Deep learning have supposed a significant breakthrough in many computer vision tasks, including facial expression analysis. Particularly, the use of Deep Convolutional Neural Networks has attained the best results in the recent public challenges. The current state-of-the-art algorithms suggest that the use of ensembles of CNNs can outperform individual CNN classifiers. Two key considerations influence these results: (i) The design of CNNs involves the adjustment of parameters that allow diversity and complementarity in the partial classification results, and (ii) the final classification rule that assembles the result of the committee. In this paper we propose to improve the assembling of the committee by introducing supervised learning on the ensemble computation. We train a CNN on the posterior-class probabilities resulting from the individual members allowing to capture non-linear dependencies among committee members, and to learn this combination from data. The validation shows an accuracy 5 percent higher with respect to previous state-of-the art results based on averaging classifiers, and 4 percent to the majority voting rule.

In order to make improvements to teaching, it is vital to know what students think of the way they are taught. With that purpose in mind, exhaustively analyzing the forums associated with the subjects taught at the Universitat Oberta de Cataluya (UOC) would be extremely helpful, as the university's students often post comments on their learning experiences in them. Exploiting the content of such forums is not a simple undertaking. The volume of data involved is very large, and performing the task manually would require a great deal of effort from lecturers. As a first step to solve this problem, we propose a tool to automatically analyze the posts in forums of communities of UOC students and teachers, with a view to systematically mining the opinions they contain. This article defines the architecture of such tool and explains how lexical-semantic and language technology resources can be used to that end. For pilot testing purposes, the tool has been used to identify students' opinions on the UOC's Business Intelligence master's degree course during the last two years. The paper discusses the results of such test. The contribution of this paper is twofold. Firstly, it demonstrates the feasibility of using natural language parsing techniques to help teachers to make decisions. Secondly, it introduces a simple tool that can be refined and adapted to a virtual environment for the purpose in question.

The amount of multimodal devices that surround us is growing everyday. In this context, human interaction and communication have become a focus of attention and a hot topic of research. A crucial element in human relations is the evaluation of individuals with respect to facial traits, what is called a first impression. Studies based on appearance have suggested that personality can be expressed by appearance and the observer may use such information to form judgments. In the context of rapid facial evaluation, certain personality traits seem to have a more pronounced effect on the relations and perceptions inside groups. The perception of dominance has been shown to be an active part of social roles at different stages of life, and even play a part in mate selection. The aim of this paper is to study to what extent this information is learnable from the point of view of computer science. Specifically we intend to determine if judgments of dominance can be learned by machine learning techniques. We implement two different descriptors in order to assess this. The first is the histogram of oriented gradients (HOG), and the second is a probabilistic appearance descriptor based on the frequencies of grouped binary tests. State of the art classification rules validate the performance of both descriptors, with respect to the prediction task. Experimental results show that machine learning techniques can predict judgments of dominance rather accurately (accuracies up to 90%) and that the HOG descriptor may characterize appropriately the information necessary for such task.

Humans constantly evaluate the personalities of other people using their faces. Facial trait judgments have been studied in the psychological field, and have been determined to influence important social outcomes of our lives, such as elections outcomes and social relationships. Recent work on textual descriptions of faces has shown that trait judgments are highly correlated. Further, behavioral studies suggest that two orthogonal dimensions, valence and dominance, can describe the basis of the human judgments from faces. In this paper, we used a corpus of behavioral data of judgments on different trait dimensions to automatically learn a trait predictor from facial pixel images. We study whether trait evaluations performed by humans can be learned using machine learning classifiers, and used later in automatic evaluations of new facial images. The experiments performed using local point-based descriptors show promising results in the evaluation of the main traits.

This article proposes a general extension of the Error Correcting Output Codes (ECOC) framework to the online learning scenario. As a result, the final classifier handles the addition of new classes independently of the base classifier used. Validation on UCI database and two real machine vision applications show that the online problem-dependent ECOC proposal provides a feasible and robust way for handling new classes using any base classifier.

This paper introduces a novel binary discriminative learning technique based on the approximation of the nonlinear decision boundary by a piecewise linear smooth additive model. The decision border is geometrically defined by means of the characterizing boundary points-points that belong to the optimal boundary under a certain notion of robustness. Based on these points, a set of locally robust linear classifiers is defined and assembled by means of a Tikhonov regularized optimization procedure in an additive model to create a final lambda-smooth decision rule. As a result, a very simple and robust classifier with a strong geometrical meaning and nonlinear behavior is obtained. The simplicity of the method allows its extension to cope with some of today's machine learning challenges, such as online learning, large-scale learning or parallelization, with linear computational complexity. We validate our approach on the UCI database, comparing with several state-of-the-art classification techniques. Finally, we apply our technique in online and large-scale scenarios and in six real-life computer vision and pattern recognition problems: gender recognition based on face images, intravascular ultrasound tissue classification, speed traffic sign detection, Chagas' disease myocardial damage severity detection, old musical scores clef classification, and action recognition using 3D accelerometer data from a wearable device. The results are promising and this paper opens a line of research that deserves further attention.

Face recognition applications commonly suffer from three main drawbacks: a reduced training set, information lying in high-dimensional subspaces, and the need to incorporate new people to recognize. In the recent literature, the extension of a face classifier in order to include new people in the model has been solved using online feature extraction techniques. The most successful approaches of those are the extensions of the principal component analysis or the linear discriminant analysis. In the current paper, a new online boosting algorithm is introduced: a face recognition method that extends a boosting-based classifier by adding new classes while avoiding the need of retraining the classifier each time a new person joins the system. The classifier is learned using the multitask learning principle where multiple verification tasks are trained together sharing the same feature space. The new classes are added taking advantage of the structure learned previously, being the addition of new classes not computationally demanding. The present proposal has been (experimentally) validated with two different facial data sets by comparing our approach with the current state-of-the-art techniques. The results show that the proposed online boosting algorithm fares better in terms of final accuracy. In addition, the global performance does not decrease drastically even when the number of classes of the base problem is multiplied by eight.

We present a method for learning discriminative linear feature extraction using independent tasks. More concretely, given a target classification task, we consider a complementary classification task that is independent of the target one. For example, in face classification field, subject recognition can be a target task while facial expression classification can be a complementary task. Then, we use labels of the complementary task in order to obtain a more robust feature extraction, being the new feature space less sensitive to the complementary classification. To learn the proposed feature extraction we use the mutual information measure between the projected data and both labels from the target and the complementary tasks. In our experiments, this framework has been applied to a face recognition problem, in order to inhibit this classification task from environmental artifacts, and to mitigate the effects of the small sample size problem. Our classification experiments show an improved feature extraction process using the proposed method.

In this paper, we propose a new supervised linear feature extraction technique for multiclass classification problems that is specially suited to the nearest neighbor classifier (NN). The problem of finding the optimal linear projection matrix is defined as a classification problem and the Adaboost algorithm is used to compute it in an iterative way. This strategy allows the introduction of a multitask learning (MTL) criterion in the method and results in a solution that makes no assumptions about the data distribution and that is specially appropriated to solve the small sample size problem. The performance of the method is illustrated by an application to the face recognition problem. The experiments show that the representation obtained following the multitask approach improves the classic feature extraction algorithms when using the NN classifier, especially when we have a few examples from each class.

In this article, we perform an extended analysis of different face-processing techniques for gender recognition problems. Prior research works show that support vector machines (SVM) achieve the best classification results. We will show that a nearest neighbor classification approach can reach a similar performance or improve the SVM results, given an adequate selection of features of the input data. This selection is performed using a dimensionality reduction technique based on a modification of nonparametric discriminant analysis, designed to improve the nearest neighbor classification. The choice of nearest neighbor is especially justified by the use of a large database. We also analyze a nonlinear algorithm, locally linear embedding, and its supervised version. Given that this technique is focused on preserving the local configuration of the neighborhood of each point, it should be a priori a good dimensionality reduction technique for extracting good features for nearest neighbor classification. A complete comparative study with the most classical face-processing techniques is also performed. © 2005 Wiley Periodicals, Inc. Int J Int Syst 20: 561–576, 2005.