Long-standing data sparsity and cold-start constitute thorny and perplexing problems for the recommendation systems. Cross-domain recommendation as a domain adaptation framework has been utilized to effectively address these challenging issues, by exploiting information from multiple domains. In this study, an item-level relevance cross-domain recommendation task is explored, where two related domains, that is, the source and the target domain contain common items. Additionally, a user-level relevance scenario is considered, where the two related domains contain common users. In light of these scenarios, two novel coupled autoencoder-based deep learning methods are proposed for cross-domain recommendation. The first method aims at simultaneously learning a pair of autoencoders in order to reveal the intrinsic representations in the source and target domains, along with a coupled mapping function to model the non-linear relationships between these representations. The second method is derived based on a new joint regularized optimization problem, which employs two autoencoders to generate in a deep and non-linear manner the user and item-latent factors, while at the same time a data-driven function is learned to map the latent factors across domains. Extensive numerical experiments are conducted illustrating the superior performance of our proposed methods compared to several state-of-the-art cross-domain recommendation frameworks.

Phishing attacks constitute a significant cyberthreat, causing financial losses and undermining digital trust. Despite advancements in awareness campaigns and regulatory efforts, many individuals and organizations struggle to understand their legal rights and obligations when victimized by phishing. This paper presents a benchmark framework designed to support researchers in the analysis of legal texts and cases, with the goal of helping users navigate the complex legal landscape surrounding phishing attacks. We evaluate BERT-based models, large language models, and retrieval-augmented generation using a curated knowledge base of phishing laws, regulations, and common scenarios. We provide baseline evaluations across multiple task categories, including true-false questions, phishing attack cases, penal code analysis, legal cases with judicial decisions, and question-answering tasks, all within the Greek legal jurisdiction.

We investigate a quantum data reduction technique with application to video classification. A hybrid quantum-classical step performs data reduction on the video dataset generating “representative” distributions for each video class. These distributions are used by a quantum classification algorithm to firstly reduce the size of the videos and then classify the reduced videos to one of$k$classes. We verify the method using sign videos and demonstrate that the reduced videos contain enough information to successfully classify them using a quantum classification process. The proposed data reduction method showcases a way to alleviate the “data loading” problem of quantum computers for the problem of video classification. Data loading is a huge bottleneck, as there are no known efficient techniques to perform that task without sacrificing many of the benefits of quantum computing.

Automating sign language translation (SLT) is a challenging real-world application. Despite its societal importance, though, research progress in the field remains rather poor. Crucially, existing methods that yield viable performance necessitate the availability of laborious to obtain gloss sequence groundtruth. In this paper, we attenuate this need, by introducing an end-to-end SLT model that does not entail explicit use of glosses; the model only needs text groundtruth. This is in stark contrast to existing end-to-end models that use gloss sequence groundtruth, either in the form of a modality that is recognized at an intermediate model stage, or in the form of a parallel output process, jointly trained with the SLT model. Our approach constitutes a Transformer network with a novel type of layers that combines: (i) local winner-takes-all (LWTA) layers with stochastic winner sampling, instead of conventional ReLU layers, (ii) stochastic weights with posterior distributions estimated via variational inference, and (iii) a weight compression technique at inference time that exploits estimated posterior variance to perform massive, almost lossless compression. We demonstrate that our approach can reach the currently best reported BLEU-4 score on the PHOENIX 2014T benchmark, but without making use of glosses for model training, and with a memory footprint reduced by more than 70%.

We present our work towards creating a dataset, which is intended to be used for the implementation of a home care services system for the deaf. The dataset includes recorded realistic scenarios of interactions between deaf patients and mental health experts in their native sign language. The scenarios allow for contextualized representations, in contrast to typical datasets presenting isolated signs or sentences. It includes continuous videos in RGB and depth, which are challenging to analyze and closely resemble real-life scenarios. The research on representation of signs is supported by providing the hand shapes and trajectories for every video using hand and skeleton models, as well as facial features. Furthermore, the dataset may be used for studying the emotional context, since such conversations are typically emotionally charged.

We present our work towards a dataset of Mycenaean Linear B sequences gathered from the Mycenaean inscriptions written in the 13th and 14th century B.C. (c. 1400-1200 B.C.). The dataset contains sequences of Mycenaean words and ideograms according to the rules of the Mycenaean Greek language in the Late Bronze Age. Our ultimate goal is to contribute to the study, reading and understanding of ancient scripts and languages. Focusing on sequences, we seek to exploit the structure of the entire language, not just the Mycenaean vocabulary, to analyse sequential patterns. We use the dataset to experiment on estimating the missing symbols in damaged inscriptions.

Decentralized computation is crucial for training on large data sets, which are stored in various locations. This paper proposes a method for collaboratively training generative adversarial networks (GANs) on several nodes to approach this problem. Each node has access to its local private dataset and model, but can influence and can be influenced by neighboring nodes. Such an approach avoids the need of a fusion center, offers better network load balance and improves data privacy. This work proposes an initial approach for decentralized GAN network optimization which is based on discriminator exchange among neighbors using an information gain criterion. The initial experiments are presented to verify whether such a decentralized architecture can provide useful results.

In this paper, we propose a discriminative framework for online simultaneous segmentation and classification of modeled visual actions that can be performed in the context of other unknown actions. To this end, we employ Hough transform to vote in a 3D space for the begin point, the end point, and the label of the segmented part of the input stream. A support vector machine is used to model each class and to suggest putative labeled segments on the timeline. To identify the most plausible segments among the putative ones, we apply a dynamic programming algorithm, which maximizes the likelihood for label assignment in linear time. The performance of our method is evaluated on synthetic as well as on real data (Weizmann, TUM Kitchen, UTKAD, and Berkeley Multimodal Human Action databases). Extensive quantitative results obtained on a number of standard data sets demonstrate that the proposed approach is of comparable accuracy with the state-of-the-art approaches for online stream segmentation and classification when all performed actions are known, and performs considerably better in the presence of unmodeled actions.

Unsupervised feature learning algorithms based on convolutional formulations of independent components analysis (ICA) have been demonstrated to yield state-of-the-art results in several action recognition benchmarks. However, existing approaches do not allow for the number of latent components (features) to be automatically inferred from the data in an unsupervised manner. This is a significant disadvantage of the state-of-the-art, as it results in considerable burden imposed on researchers and practitioners, who must resort to tedious cross-validation procedures to obtain the optimal number of latent features. To resolve these issues, in this paper we introduce a convolutional nonparametric Bayesian sparse ICA architecture for overcomplete feature learning from high-dimensional data. Our method utilizes an Indian buffet process prior to facilitate inference of the appropriate number of latent features under a hybrid variational inference algorithm, scalable to massive datasets. As we show, our model can be naturally used to obtain deep unsupervised hierarchical feature extractors, by greedily stacking successive model layers, similar to existing approaches. In addition, inference for this model is completely heuristics-free, thus, it obviates the need of tedious parameter tuning, which is a major challenge most deep learning approaches are faced with. We evaluate our method on several action recognition benchmarks, and exhibit its advantages over the state-of-the-art.

In this paper, we propose a Gaussian process (GP) model for analysis of nonlinear time series. Formulation of our model is based on the consideration that the observed data are functions of latent variables, with the associated mapping between observations and latent representations modeled through GP priors. In addition, to capture the temporal dynamics in the modeled data, we assume that subsequent latent representations depend on each other on the basis of a hidden Markov prior imposed over them. Derivation of our model is performed by marginalizing out the model parameters in closed form using GP priors for observation mappings, and appropriate stick-breaking priors for the latent variable (Markovian) dynamics. This way, we eventually obtain a nonparametric Bayesian model for dynamical systems that accounts for uncertainty in the modeled data. We provide efficient inference algorithms for our model on the basis of a truncated variational Bayesian approximation. We demonstrate the efficacy of our approach considering a number of applications dealing with real-world data, and compare it with the related state-of-the-art approaches.

In this paper, we propose a novel system for visual recognition and summarization of pick and place tasks that may be executed in settings such as an industrial assembly line. Our novel approach is based on the utilization of hidden Markov models for online task recognition as well as on the use of prior knowledge via a Hopfield-based optimization scheme. To facilitate offline analysis, we extract summaries of the captured content based on these features. We extract the motion energy using the norms of the Zernike moments, looking for local minima and maxima that indicate distinctive visual events and thus key-frames. The proposed scheme is not threshold-dependent, and, therefore, the number of extracted key-frames varies according to the complexity of motion energy variation. We validate our system by experimenting on two datasets.

In this paper, we provide a variational Bayesian (VB) treatment of multistream fused hidden Markov models (MFHMMs), and apply it in the context of active learning-based visual workflow recognition (WR). Contrary to training methods yielding point estimates, such as maximum likelihood or maximum a posteriori training, the VB approach provides an estimate of the posterior distribution over the MFHMM parameters. As a result, our approach provides an elegant solution toward the amelioration of the overfitting issues of point estimate-based methods. Additionally, it provides a measure of confidence in the accuracy of the learned model, thus allowing for the easy and cost-effective utilization of active learning in the context of MFHMMs. Two alternative active learning algorithms are considered in this paper: query by committee, which selects unlabeled data that minimize the classification variance, and a maximum information gain method that aims to maximize the alteration in model variance by proper data labeling. We demonstrate the efficacy of the proposed treatment of MFHMMs by examining two challenging WR scenarios, and show that the application of active learning, which is facilitated by our VB approach, allows for a significant reduction of the MFHMM training costs.

In this paper we introduce the WR (Workflow Recognition) dataset. Recorded in the production line of a major automobile manufacturer, this dataset consists of sequences that depict workers executing industrial workflows. The heavy occlusions, outliers, the visually complicated background and the human-machinery interaction are among the factors that make this dataset a very challenging testbed for computer vision and image processing algorithms. We provide the original video sequences together with event labeling, as well as feature vectors extracted through our proposed scene representation methodology, and we refer to our results so far in workflow recognition using this dataset.

In this work, we propose a framework for classifying structured human behavior in complex real environments, where problems such as frequent illumination changes and heavy occlusions are expected. Since target recognition and tracking can be very challenging, we bypass these problems by employing an approach similar to Motion History Images for feature extraction. Furthermore, to tackle outliers residing within the training data, which might affect severely the training algorithm of models with Gaussian observation likelihoods, we scrutinize the effectiveness of the multivariate Student-t distribution as the observation likelihood of the employed Hidden Markov Models. Additionally, the problem of visibility and occlusions is addressed by providing various extensions of the framework for multiple cameras, both at the feature and at the state level. Finally, we evaluate the performance of the examined approaches under real-life visual behavior understanding scenarios and we compare and discuss the obtained results.

Hidden Markov (chain) models using finite Gaussian mixture models as their hidden state distributions have been successfully applied in sequential data modeling and classification applications. Nevertheless, Gaussian mixture models are well known to be highly intolerant to the presence of untypical data within the fitting data sets used for their estimation. Finite Student's t-mixture models have recently emerged as a heavier-tailed, robust alternative to Gaussian mixture models, overcoming these hurdles. To exploit these merits of Student's t-mixture models in the context of a sequential data modeling setting, we introduce, in this paper, a novel hidden Markov model where the hidden state distributions are considered to be finite mixtures of multivariate Student's t-densities. We derive an algorithm for the model parameters estimation under a maximum likelihood framework, assuming full, diagonal, and factor-analyzed covariance matrices. The advantages of the proposed model over conventional approaches are experimentally demonstrated through a series of sequential data modeling applications.

Hidden Markov models using finite Gaussian mixture models as their hidden state distributions have been applied in modeling of time series that result from various noisy signals. Nevertheless, Gaussian mixture models are well-known to be highly intolerant to the presence of outliers within the fitting sets used for their estimation. Finite Student's-i mixture models have recently emerged as a heavier-tailed, robust alternative to Gaussian mixture models, overcoming these hurdles. To exploit those merits of Student's-i mixture models, we introduce in this paper a novel hidden Markov chain model where the hidden state distributions are considered to be finite mixtures of multivariate Student's-i densities and we derive an algorithm for the model parameters estimation under a maximum likelihood framework. We apply this novel approach in automatic gesture recognition and we show that our model provides a substantial improvement in data representation performance and computational efficiency over the standard Gaussian model.

The high potential of Superquadrics as modeling elements for image segmentation tasks has been pointed out since years in the computer vision community. In this work we employ superquadrics as modeling elements for multiple object segmentation in range images. Segmentation is executed in two stages. Firstly, a hypothesis about the values of the segmentation parameters is generated. Secondly, the hypothesis is refined locally. In both stages, object boundary and region information are considered. Boundary information is derived via model-based edge detection in the input range image. Hypothesis generation uses boundary information to isolate image regions which can be accurately described by superquadrics. Within hypothesis refinement, a game-theoretic framework is used to fuse the two information sources by associating an objective function to each information source. Iterative optimization of the two objective functions in succession, outputs a precise description of all image objects. We demonstrate experimentally that this approach substantially improves the most established method in superquadric segmentation, in terms of accuracy and computational efficiency. We demonstrate the applicability of our segmentation framework in real world applications by constructing a novel robotic system for automatic unloading of jumbled box-like objects from platforms.

A new method for object tracking in video sequences is presented. This method exploits the benefits of particle filters to tackle the multimodal distributions emerging from cluttered scenes. The tracked object is described by several models of different complexity, which are probabilistically linked together. The parameter update for each model takes place hierarchically so that the simpler models, which are updated first, can guide the search in the parameter space of the more complex models to relevant regions. This strategy improves the target representation because of the multiple models and reduces the overall complexity. The likelihood for each object model is calculated using one or more visual cues thus increasing the robustness of the proposed algorithm. Our method is evaluated by fusing on salient points and contour models and we demonstrate its effectiveness.

A novel method for summarizing videos of gestures is presented. The gestures performed by the hands and the head are extracted through skin color segmentation and represented through Zernike moments. The gesture energy is calculated using the norms of the Zernike moments and monitored through time for local minima and maxima that indicate distinctive visual events and thus key-frames. The proposed scheme is not threshold-dependent and therefore the number of extracted key-frames varies according to the complexity of gesture energy variation. The applicability of the method is verified experimentally in sign language videos.

We propose an efficient approach towards the solution of the de-palletizing problem, based on active vision. We describe a system comprising an industrial robot and a time of flight laser sensor, which performs the depalletizing task in real time, and independently of lighting conditions. In our case, the target objects are solid boxes of known identical dimensions, neatly layered but with arbitrary orientation within a layer, which are all placed on a platform. The layered structure of the target platform allows for 2D imagery. The system locates the position of the boxes by tracking one of the corners they expose to the laser source. The system locates the desired corners by applying the scan line approximation technique, adapted to fit the needs of our application, to the 2D input data. The advantages of our system over existing applications are its simplicity, robustness, speed and ease of installation.