@article {166, title = {Cost-efficient coupled learning methods for recovering near-infrared information from RGB signals: Application in precision agriculture}, journal = {Computers and Electronics in Agriculture}, volume = {209}, year = {2023}, abstract = {[size= 16px; box-sizing: border-box; margin: 0px; padding: 0px; caret-color: $\#$2e2e2e; color: $\#$2e2e2e; font-family: ElsevierGulliver, Georgia, {\textquoteright}Times New Roman{\textquoteright}, STIXGeneral, {\textquoteright}Cambria Math{\textquoteright}, Arial, Helvetica, {\textquoteright}Lucida Sans Unicode{\textquoteright}, {\textquoteright}Microsoft Sans Serif{\textquoteright}, {\textquoteright}Segoe UI Symbol{\textquoteright}, {\textquoteright}Arial Unicode MS{\textquoteright}, serif][url=https://www.sciencedirect.com/topics/computer-science/multispectral-imaging]Multispectral imaging[/url] and the derived [/size][url=https://www.sciencedirect.com/topics/agricultural-and-biological-sciences/spectral-analysis]spectral analysis[/url] offer useful tools for revealing beneficial information for a variety of applications, e.g., precision agriculture, [url=https://www.sciencedirect.com/topics/computer-science/medical-imaging]medical imaging[/url] and [url=https://www.sciencedirect.com/topics/computer-science/autonomous-driving]autonomous driving[/url]. Contrary to mainstream RGB cameras that can capture information derived only from three bands within the visible spectrum, the multispectral cameras can offer better spectral resolution by utilizing the underlying information in the visible and the near-infrared spectrum. However, the cost of the multispectral cameras is very high and their mobility is limited due to their weight and their need for special hardware equipment. Considering the aforementioned limitations, we propose two low-cost and efficient methods to infer detailed [url=https://www.sciencedirect.com/topics/computer-science/spectral-information]spectral information[/url] outside the visible spectrum range by employing only an RGB camera. The proposed methods require significantly less training data, containing approximately [size= 16px; box-sizing: border-box; margin: 0px; padding: 0px; caret-color: $\#$2e2e2e; color: $\#$2e2e2e; font-family: ElsevierGulliver, Georgia, {\textquoteright}Times New Roman{\textquoteright}, STIXGeneral, {\textquoteright}Cambria Math{\textquoteright}, Arial, Helvetica, {\textquoteright}Lucida Sans Unicode{\textquoteright}, {\textquoteright}Microsoft Sans Serif{\textquoteright}, {\textquoteright}Segoe UI Symbol{\textquoteright}, {\textquoteright}Arial Unicode MS{\textquoteright}, serif]99.8\%[/size][size= 16px; box-sizing: border-box; margin: 0px; padding: 0px; caret-color: $\#$2e2e2e; color: $\#$2e2e2e; font-family: ElsevierGulliver, Georgia, {\textquoteright}Times New Roman{\textquoteright}, STIXGeneral, {\textquoteright}Cambria Math{\textquoteright}, Arial, Helvetica, {\textquoteright}Lucida Sans Unicode{\textquoteright}, {\textquoteright}Microsoft Sans Serif{\textquoteright}, {\textquoteright}Segoe UI Symbol{\textquoteright}, {\textquoteright}Arial Unicode MS{\textquoteright}, serif] less parameters compared to the competing [url=https://www.sciencedirect.com/topics/computer-science/deep-learning]deep learning[/url] approaches and can be deployed on various edge devices with computational and power constraints, e.g., mobile phones or unmanned drones for addressing problems in precision agriculture under real-field settings. Extensive numerical results demonstrate the efficacy of the proposed models to reconstruct images outside the visible spectrum. Additionally, the reconstructed images can be utilized to estimate the [url=https://www.sciencedirect.com/topics/agricultural-and-biological-sciences/normalized-difference-vegetation-index]normalized difference vegetation index[/url] (NDVI), which can reveal valuable information concerning the health of the monitored plants without the need of a multispectral camera.[/size]}, author = {Alexandros Gkillas and Dimitrios Kosmopoulos and Kostas Berberidis} } @proceedings {168, title = {Developing Robust and Lightweight Adversarial Defenders by Enforcing Orthogonality on Attack-Agnostic Denoising Autoencoders}, journal = {International Conference on Computer Vision, Workshop on Resource Efficient Deep Learning for Computer Vision}, year = {2023}, abstract = {Adversarial attacks have become a critical threat to the security and reliability of machine learning models. We propose a solution to the problem of defending against adversarial attacks using a deep Denoising Auto Encoder (DAE). The proposed DAE is trained to enforce orthogonality between the noise and the range space of its output in each layer of the encoder{\textquoteright}s chain. Furthermore, the pseudoinverse decoder of the DAE is designed to ensure that the reconstructed image and the null space of its intermediate representations in each layer of the chain maintain orthogonality as it progresses from the target space to the latent space. The denoising problem is formulated as an equality constrained optimization problem, which is solved by finding the stationary points of the Lagrangian function. The noisy data are generated by adding realizations of multiple random noise distributions to pristine data during DAE training, resulting in excellent denoising performance. We compare the performance of our full weights and tied-weights DAEs, showing that the latter not only has half the complexity of the former, but also outperforms the former in denoising and in strong adversarial attacks. To demonstrate the effectiveness of the proposed solution we evaluate our networks against recent works in the literature, specifically those focusing on defending against adversarial attacks.}, author = {Aristeidis Bifis and Emmanuel Psarakis and Dimitrios Kosmopoulos} } @article {162, title = {A generative model for the Mycenaean Linear B script and its application in infilling text from ancient tablets}, journal = {ACM Journal on Computing and Cultural Heritage}, volume = {16}, year = {2023}, pages = {1-25}, abstract = {We present a generative neural language model for the most ancient proven stage of the Greek language, the Mycenaean Greek attributed by the Linear B script. To capture the statistical structure of the Mycenaean documents we present a Bidirectional Recurrent Neural Network and compare it to the traditionally used n-grams. The model is used to supplement the damaged parts of the Mycenaean texts, namely the incomplete, to a greater or lesser extent, words, which are typically discovered on partially damaged clay tablets. We verify our method experimentally using ground - truth, then we demonstrate our results on real cases and compare with experts{\textquoteright} opinions. We also present a methodology to augment our dataset, which turns out to improve our results.}, author = {K. PAPAVASSILEIOU and Dimitrios I. Kosmopoulos} } @conference {167, title = {Procrustes-DTW: Dynamic Time Warping Variant for the Recognition of Sign Language Utterances}, booktitle = {2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW)}, year = {2023}, abstract = {[color=$\#$333333; font-family: {\textquoteright}HelveticaNeue Regular{\textquoteright}, sans-serif; font-size: 18px; font-variant-ligatures: normal; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial]In this paper we present a method for classifying sign language videos using a variation of dynamic time warping with the Procrustes distance, which is a shape similarity measure appropriate for handshape recognition. We initially extract features from the dominant hand. We then classify the signs using a nearest-neighbor scheme which includes a dynamic time warping variant, which we dub Procrustes-DTW and which is suitable for extracting a similarity measure of sequences of handshapes only. To reduce computational costs it is combined with a curvature-based summarization scheme with a reasonable sacrifice in accuracy. We verify experimentally our approach on a custom dataset and we show its merit compared to the conventional dynamic time warping approach.[/color]}, doi = {10.1109/ICASSPW59220.2023.10193012}, author = {Arvanitis, Nikolaos and Sartinas, Evangelos and Kosmopoulos, Dimitrios} } @article {156, title = {Context-Aware Automatic Sign Language Video Transcription in Psychiatric Interviews}, journal = {Sensors}, volume = {22}, year = {2022}, pages = {2656}, abstract = {Sign language (SL) translation constitutes an extremely challenging task when undertaken in a general unconstrained setup, especially in the absence of vast training datasets that enable the use of end-to-end solutions employing deep architectures. In such cases, the ability to incorporate prior information can yield a significant improvement in the translation results by greatly restricting the search space of the potential solutions. In this work, we treat the translation problem in the limited confines of psychiatric interviews involving doctor-patient diagnostic sessions for deaf and hard of hearing patients with mental health problems.To overcome the lack of extensive training data and be able to improve the obtained translation performance, we follow a domain-specific approach combining data-driven feature extraction with the incorporation of prior information drawn from the available domain knowledge. This knowledge enables us to model the context of the interviews by using an appropriately defined hierarchical ontology for the contained dialogue, allowing for the classification of the current state of the interview, based on the doctor{\textquoteright}s question. Utilizing this information, video transcription is treated as a sentence retrieval problem. The goal is predicting the patient{\textquoteright}s sentence that has been signed in the SL video based on the available pool of possible responses, given the context of the current exchange. Our experimental evaluation using simulated scenarios of psychiatric interviews demonstrate the significant gains of incorporating context awareness in the system{\textquoteright}s decisions.}, author = {Erion-Vasilis Pikoulis and Aristeidis Bifis and Maria Trigka and Constantinos Constantinopoulos and Dimitrios Kosmopoulos} } @proceedings {155, title = {Exploitation of Noisy Automatic Data Annotation and Its Application to Hand Posture Classification}, journal = {International Joint Conference on Computer Vision Imaging and Computer Graphics Theory and Applications}, year = {2022}, abstract = {The success of deep learning in recent years relies on the availability of large amounts of accurately annotated training data. In this work, we investigate a technique for utilizing automatically annotated data in classification problems. Using a small number of manually annotated samples, and a large set of data that feature automatically created, noisy labels, our approach trains a Convolutional Neural Network (CNN) in an iterative manner. The automatic annotations are combined with the predictions of the network in order to gradually expand the training set. In order to evaluate the performance of the proposed approach, we apply it to the problem of hand posture recognition from RGB images. We compare the results of training a CNN classifier with and without the use of our technique. Our method yields a significant increase in average classification accuracy, and also decreases the deviation in class accuracies, thus indicating the validity and the usefulness of the proposed approach.}, author = {Georgios Lydakis and Iason Oikonomidis and Dimitrios Kosmopoulos and Antonis Argyros} } @article {164, title = {A Multispectral Dataset for the Detection of Tuta absoluta and Leveillula taurica in Tomato Plants}, journal = {Smart Agricultural Technology}, year = {2022}, pages = {100146}, abstract = {Tomato (Solanum lycopersicum) is one of the most important vegetables for human nutrition and its cultivation employs amounts of resources worldwide. However, tomato cultivation is plagued by several diseases and pests that increase production cost and introduce additional environmental and health risks due to pesticide use. Timely disease and pest detection is of high importance for tomato crop output and the environment, since plant protection input can be optimized. Here, we present a dataset of multispectral images (RGB and NIR) of tomato plants, at various stages of infection with Tuta absoluta and Leveillula taurica, which to our knowledge is unique. The dataset comprised of 263 images collected from a real greenhouse. Additionally, we applied a baseline Faster-RCNN object detector for the localization and classification lesions. Our experiments include (i) a version for the RGB channels and (ii) a custom backbone architecture version for feature fusion using the same Faster-RCNN head. Lastly, based on the detector{\textquoteright}s output, we compute an \>0.9 F1-score on binary classification, while mAP 18.5\% and mAP 20.2\% on detection, highlight the added value of NIR spectral bands for detecting these diseases.}, keywords = {Faster-RCNN, Feature fusion, Leaf Miner, Plant Disease Dataset, Powdery Mildew}, issn = {2772-3755}, doi = {https://doi.org/10.1016/j.atech.2022.100146}, url = {https://www.sciencedirect.com/science/article/pii/S2772375522001101}, author = {P.S. Georgantopoulos and D. Papadimitriou and C. Constantinopoulos and T. Manios and I.N. Daliakopoulos and D. Kosmopoulos} } @proceedings {163, title = {Quantum Data Reduction with Application to Video Classification}, journal = {ACM/IEEE Workshop on Quantum Computing}, year = {2022}, abstract = {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{\textquoteright}{\textquoteright} 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 {\textquoteleft}{\textquoteleft}data loading{\textquoteright}{\textquoteright} 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.}, author = {Kostas Blekos and Dimitrios Kosmopoulos} } @proceedings {153, title = {A 2-D Wrist Motion Based Sign Language Video Summarization}, journal = {British Machine Vision Conference - ORAL PRESENTATION}, year = {2021}, abstract = {In this paper we present a keyframe extraction scheme based on the wrist motion using differential geometry. More specifically, the time (t)-parameterized Frennet-Serret frame for tracking the signer{\textquoteright}s wrist is used and the curvature of the trajectory, is proposed for the identification of the Sign Language (SL) video keyframes. Specifically, a video frame is characterized as keyframe if on that time instance the t-parameterized curvature function attains a maximum value. Finally, in order to properly define the wrist 2-D motion model, a skeleton tracker is used. The proposed scheme is adaptable, i.e., the number of extracted keyframes varies according to the complexity of the signs, while preserving the semantic content. This in turn makes it attractive for applications like video-calling. Its performance in terms of the achieved compression and intelligibility ratios was evaluated on a ground-truth sequence and outperformed its s-parameterized counterpart (s is the arc length); it also outperformed a moment-based SL summarization technique. Furthermore, the proposed scheme was experimentally evaluated on a dataset containing 5500 signs by SL specialists with very promising results. Finally, the proposed keyframe extraction was evaluated against the aforementioned techniques on the same dataset via the use of a GRU neural network on the gloss classification problem; its superior accuracy in identifying the gloss meaning was confirmed.}, author = {Evangelos Sartinas and Emmanouil Psarakis and Klimis Antzakas and Dimitrios Kosmopoulos} } @article {154, title = {A cross-domain recommender system using deep coupled autoencoders}, year = {2021}, month = {8 December 2021}, abstract = {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 efficiently 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 without sharing sensitive information regarding the users{\textquoteright} behavior, and thus avoiding the leak of user privacy. In light of this scenario, two novel coupled autoencoder-based deep learning methods are proposed for cross-domain recommendation. The first method aims to simultaneously learn a pair of autoencoders in order to reveal the intrinsic representations of the items in the source and target domains, along with a coupled mapping function to model the non-linear relationships between these representations, thus transferring beneficial information from the source to the target domain. 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 learnt to map the item-latent factors across domains. Extensive numerical experiments on two publicly available benchmark datasets are conducted illustrating the superior performance of our proposed methods compared to several state-of-the-art cross-domain recommendation frameworks.}, author = {Alexandros Gkillas and Dimitrios Kosmopoulos} } @proceedings {147, title = {A Method for Recovering Near Infrared Information from RGB Measurements with Application in Precision Agriculture}, journal = {European Signal Processing Conference}, year = {2021}, pages = {616-620}, publisher = {IEEE}, edition = {IEEE}, abstract = {[size= 13.008px]In this work we develop a cost-efficient coupled dictionary learning based method for reconstructing multispectral images using only a single RGB commercial camera, without requiring the sensitivity function of the camera sensor. Considering the very high cost, the acquisition time and reduced mobility of multispectral cameras we claim that this is a very attractive option. In contrast to other approaches, the proposed method is not limited only to spectral bands inside the visible spectrum, but it also considers an even more challenging task, that is the reconstruction of spectral bands outside the visible range closer to the near-infrared wavelengths of the spectrum. Extensive experiments with real data demonstrate the effectiveness and applicability of the proposed method in the precision agriculture domain. To this end, we calculate one of the most widely used vegetation indices, the normalized difference vegetation index (NVDI), which may be used for plant health monitoring.[/size]}, author = {A. Gkillas and D. Kosmopoulos and C. Constantinopoulos and D. Ampeliotis and K. Berberidis} } @proceedings {146, title = {PupilTAN: A Few-Shot Adversarial Pupil Localizer}, journal = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, GAZE 2021: Gaze Estimation and Prediction in the Wild }, year = {2021}, pages = {3134-3142}, publisher = {BEST PAPER AWARD}, abstract = {
The eye center localization is a challenging problem faced by many computer vision applications. The challenges typically stem from the scene variability, such as, the wide range of shapes, the lighting conditions, the view angles and the occlusions. Nowadays, the increasing interest on deep neural networks requires a large volume of training data. However, a significant issue is the dependency on labeled data, which are expensive to obtain and susceptible to errors. To address these issues, we propose a deep network, dubbed PupilTAN, that performs image to-heatmap Translation and an Adversarial training framework that solves the eye localization problem in a few-shot unsupervised way. The key idea is to estimate, by using only a few ground-truth shots, the heatmaps centers{\textquoteright} pdf and use it as a generator to create random heatmaps that follow the same probability distribution of the real ones. We showcase that training the deep network with these artificial heatmaps in an adversarial framework not only makes us less dependent on labeled data, but also leads to a significant accuracy improvement. The proposed network achieves realtime performance in a general-purpose computer environment and improves the state-of-the-art accuracy for both MUCT and BioID datasets, even compared with supervised techniques. Furthermore, our model is robust even in the case of reducing its size of up to 1/16 of the original network (0.2M parameters), demonstrating comparable accuracy to the state-of-the-art with high practical value to real-time applications.
}, author = {Nikolaos Poulopoulos and Emmanouil Z. Psarakis and Dimitrios Kosmopoulos} } @proceedings {151, title = {A Quantum 3D Convolutional Neural Network with Application in Video Classification}, journal = {International Symposium on Visual Computing}, year = {2021}, pages = {601-612}, edition = {G. Bebis et al (Eds.)}, abstract = {Quantum computing seeks to exploit the properties of quantum mechanics to perform computations at a fraction of the cost compared to the classical computing methods. Recently, quantum methods for machine learning have attracted the interest of researchers. Those methods aim to exploit, in the context of machine learning, the potential benefits that the quantum computers should be able to offer in the near future. A particularly interesting area of research in this direction, investigates the union of quantum machine learning models with Convolutional Neural Networks. In this paper we develop a quantum counterpart of a 3D Convolutional Neural Network for video classification, dubbed Q3D-CNN. This is the first approach for quantum video classification we are aware of. Our model is based on previously proposed quantum machine learning models, where manipulation of the input data is performed in such a way that a fully quantum-mechanical neural network layer can be realized and used to form a Quantum Convolutional Neural Network. We augment this approach by introducing quantum-friendly operations during data-loading and appropriately manipulating the quantum network. We demonstrate the applicability of the proposed Q3D-CNN in video classification using videos from a publicly available dataset. We successfully classify the test dataset using two and three classes using the quantum network and its classical counterpart.}, author = {Konstandinos Blekos and Dimitrios Kosmopoulos} } @proceedings {149, title = {Stochastic Transformer Networks with Linear Competing Units: Application to end-to-end SL Translation}, journal = {Proceedings of the IEEE/CVF International Conference on Computer Vision}, year = {2021}, pages = {11946-11955}, abstract = {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\%.}, author = {Andreas Voskou and Konstantinos Panousis and Dimitrios Kosmopoulos and Dimitris Metaxas and Sotirios Chatzis} } @proceedings {135, title = {A Dataset of Mycenaean Linear B Sequences}, journal = {Language Resources and Evaluation Conference - LREC}, year = {2020}, pages = {2552-2561}, abstract = {[size= 13.008px]We present 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 present an initial experiment on estimating the missing symbols in damaged inscriptions using the dataset.[/size]}, author = {K. Papavasileiou and G. Owens and D. Kosmopoulos} } @proceedings {134, title = {Towards a visual Sign Language dataset for home care services}, journal = {15th IEEE International Conference on Face and Gesture Recognition}, volume = {1}, year = {2020}, month = {2020}, pages = {622-626}, abstract = {[size= 13.008px]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 the study of emotional context in Sign Language, since such conversations are typically emotionally charged. [/size]}, author = {D. Kosmopoulos and I. Oikonomidis and K. Konstantinopoulos and N. Arvanitis and K. Antzakas and A. Bifis and G. Lydakis and A. Roussos and A. Argyros} } @proceedings {145, title = {Variational Bayesian Sequence-to-Sequence Networks for Memory-Efficient Sign Language Translation}, journal = {International Symposium on Visual Computing}, year = {2020}, pages = {251-262}, publisher = {Springer International Publishing}, address = {Cham}, abstract = {Memory-efficient continuous Sign Language Translation is a significant challenge for the development of assisted technologies with real-time applicability for the deaf. In this work, we introduce a paradigm of designing recurrent deep networks whereby the output of the recurrent layer is derived from appropriate arguments from nonparametric statistics. A novel variational Bayesian sequence-to-sequence network architecture is proposed that consists of a) a full Gaussian posterior distribution for data-driven memory compression and b) a nonparametric Indian Buffet Process prior for regularization applied on the Gated Recurrent Unit non-gate weights. We dub our approach Stick-Breaking Recurrent network and show that it can achieve a substantial weight compression without diminishing modeling performance.}, isbn = {978-3-030-64559-5}, author = {Partaourides, Harris and Voskou, Andreas and Kosmopoulos, Dimitrios and Chatzis, Sotirios and N. Metaxas, Dimitris} } @proceedings {133, title = {Translation of Sign Language Glosses to Text Using Sequence-to-Sequence Attention Models}, journal = {The 15th International Conference on Signal Image Technology \& Internet based Systems - SITIS}, year = {2019}, month = {2019}, abstract = {This work deals with the problem of Sign Language Translation and more specifically with translating Glosses to text. We applied Sequence to Sequence models with attention mechanism to a parallel gloss to English corpus. This is the first work that used these models to translate American gloss sentences to English. We present our experiments on several network architectures with three different attention functions. The results are very promising and can be useful for the further implementation of a full sign language recognition system.}, author = {N. Arvanitis and C. Constantinopoulos and D. Kosmopoulos} } @article {132, title = {Facilitation of Air Traffic Control via OCR-based Aircraft Registration Number Extraction}, journal = {IET Intelligent Transportation Systems}, volume = {12}, year = {2018}, pages = {965-975}, abstract = {To identify any aircraft in the world, it is sufficient to read its registration number. This number is a unique identifier, and offers valuable information, in the same way a car registration number does. In this work we present the results of our feasibility study towards a simple, yet very efficient and effective system to identify aircrafts using video-Optical Character Recognition acquired by off-the-shelf cameras. We used several videos under realistic conditions at the Heraklion airport during high-season and we achieved very promising results. We claim that there is much room for the development of a low-cost airport surface monitoring system based on standard cameras, which can complement high-cost radars.
}, author = {Dimitrios G. Vidakis and Dimitrios I. Kosmopoulos} } @proceedings {130, title = {A Prototype Towards Modeling Visual Data Using Decentralized Generative Adversarial Networks}, journal = {IEEE International Conference on Image Processing}, year = {2018}, month = {in press}, abstract = {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, higher robustness to network errors 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. We present our initial experiments to verify whether such a decentralized architecture can provide useful results.
}, author = {Dimitrios I. Kosmopoulos} } @article {131, title = {A Survey on Developing Personalized Content Services in Museums}, journal = {Pervasive and Mobile Computing}, volume = {47}, year = {2018}, pages = {54-77}, abstract = {The personalized content services in museums are motivated by the need to enhance the visitors{\textquoteright} experience through recommendations which consider the context of their visit, and by the need of curators to measure objectively the exhibition{\textquoteright}s impact. We survey the latest advancements in the fields of indoor localization, visitor
profiling, content storage and presentation, as well as curator visualization tools, which are the main elements of such systems, and we highlight their strengths and weaknesses. We present an information architecture, whichmay offer useful insights to researchers and developers. Finally, we present the current challenges and the future
trends.
}, author = {Dimitrios I. Kosmopoulos and George Styliaras} } @article {129, title = {A framework for online segmentation and classification of modeled actions performed in the context of unmodeled ones}, journal = {IEEE Transactions on Circuits and Systems for Video Technology}, volume = {27}, year = {2017}, pages = {2578-2590}, abstract = {In this work, 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. An SVM 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 datasets demonstrate that the proposed approach is of comparable accuracy to the state of the art for online stream segmentation and classification when all performed actions are known and performs considerably better in the presence of unmodeled actions.}, author = {Dimitrios Kosmopoulos and Konstantinos Papoutsakis and Antonis Argyros} } @article {122, title = {A Latent Manifold Markovian Dynamics Gaussian Process}, journal = {Neural Networks and Learning Systems, IEEE Transactions on}, volume = {26}, year = {2015}, month = {01/2015}, pages = {70-83}, abstract = {Hidden Markov models (HMMs) are a popular approach for modeling sequential data, typically based on the assumption of a first-order Markov chain. In other words, only one-step back dependencies are modeled which is a rather unrealistic assumption in most applications. In this paper, we propose a method for postulating HMMs with approximately infinitely-long time-dependencies. Our approach considers the whole history of model states in the postulated dependencies, by making use of a recently proposed nonparametric Bayesian method for modeling label sequences with infinitely-long time dependencies, namely the sequence memoizer. We manage to derive training and inference algorithms for our model with computational costs identical to simple first-order HMMs, despite its entailed infinitely-long time-dependencies, by employing a mean-field-like approximation.The efficacy of our proposed model is experimentally demonstrated.
}, author = {Sotirios Chatzis and Dimitrios Kosmopoulos and George Papadourakis} } @proceedings {126, title = {A Non-Stationary Infinite Partially Observable Markov Decision Process}, journal = {International Conference on Artificial Neural Networks}, year = {2014}, abstract = {Partially Observable Markov Decision Processes (POMDPs) have been met with great success in planning domains where agents must balance actions that provide knowledge and actions that provide reward. Recently, nonparametric Bayesian methods have been successfully applied to POMDPs to obviate the need of a priori knowledge of the size of the state space, allowing to assume that the number of visited states may grow as the agent explores its environment. These approaches rely on the assumption that the agent{\textquoteright}s environment remains stationary; however, in real-world scenarios the environment may change over time. In this work, we aim to address this inadequacy by introducing a dynamic nonparametric Bayesian POMDP model that both allows for automatic inference of the (distributional) representations of POMDP states, and for capturing non-stationarity in the modeled environments. Formulation of our method is based on imposition of a suitable dynamic hierarchical Dirichlet process (dHDP) prior over state transitions. We derive efficient algorithms for model inference and action planning and evaluate it on several benchmark tasks.}, author = {Sotirios P. Chatzis and Dimitrios I. Kosmopoulos} } @proceedings {125, title = {Online segmentation and classification of modeled actions performed in the context of unmodeled ones}, journal = {British Machine Vision Conference}, year = {2014}, abstract = {In this work, we provide a discriminative framework for online simultaneous segmentation and classification of visual actions, which deals effectively with unknown sequences that may interrupt the known sequential patterns. 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. An SVM 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 maximises an objective function for label assignment in linear time. The performance of our method is evaluated on synthetic as well as on real data (Weizmann and Berkeley multimodal human action database). The proposed approach is of comparable accuracy to the state of the art for online stream segmentation and classification and performs considerably better in the presence of previously unseen actions.}, author = {Dimitrios Kosmopoulos and Kostas Papoutsakis and Antonis Argyros} } @inbook {124, title = {Plant Leaf Recognition Using Zernike Moments and Histogram of Oriented Gradients}, booktitle = {Artificial Intelligence: Methods and Applications}, series = {Lecture Notes in Computer Science}, volume = {8445}, year = {2014}, pages = {406-417}, publisher = {Springer International Publishing}, organization = {Springer International Publishing}, abstract = {A method using Zernike Moments and Histogram of Oriented Gradients for classification of plant leaf images is proposed in this paper. After preprocessing, we compute the shape features of a leaf using Zernike Moments and texture features using Histogram of Oriented Gradients and then the Sup-port Vector Machine classifier is used for plant leaf image classification and recognition. Experimental results show that using both Zernike Moments and Histogram of Oriented Gradients to classify and recognize plant leaf image yields accuracy that is comparable or better than the state of the art. The method has been validated on the Flavia and the Swedish Leaves datasets as well as on a combined dataset.}, keywords = {Histogram of oriented gradients, Leaf recognition, Support vector machine, Zernike moments}, isbn = {978-3-319-07063-6}, doi = {10.1007/978-3-319-07064-3_33}, url = {http://dx.doi.org/10.1007/978-3-319-07064-3_33}, author = {Tsolakidis, DimitrisG. and Kosmopoulos, DimitriosI. and Papadourakis, George}, editor = {Likas, Aristidis and Blekas, Konstantinos and Kalles, Dimitris} } @article {123, title = {A top-down event-driven approach for concurrent activity recognition}, journal = {Multimedia Tools Appl.}, volume = {69}, year = {2014}, pages = {293-311}, abstract = {In this paper a framework for automatic online workflow recognition in industrial environments where the issue of concurrent activities rises, is presented. The framework consists of three main parts: The first part is devoted to detecting activity in specific Regions of Interest (ROIs) of the video sequence. This is effected by separating each frame into ROIs and representing the resulting subimages through feature vectors. By observing these vectors we can determine when there is action in a particular ROI. The second part of the framework lies in examining whether the detected activity corresponds to a workflow related event. This is accomplished by HMM modeling. Finally, the third part employs a string matching based technique to confirm the validity of the observed sequence of events or correct any detection or classification errors. This last step also addresses a top down approach by informing lower system levels (such as image representation or object tracking) about the errors committed. The performance of the proposed approach is thoroughly evaluated under real-life complex visual workflow understanding scenarios, in an industrial plant. The obtained results are compared and discussed.
}, keywords = {Concurrent activity recognition Event detection Top-down Region of interest (ROI) Workflow}, author = {Athanasios Voulodimos and Dimitrios I. Kosmopoulos and Nikolaos D. Doulamis and Theodora A. Varvarigou} } @article {38, title = {A Conditional Random Field-Based Model for Joint Sequence Segmentation and Classification}, journal = {Pattern Recognition}, volume = {46}, year = {2013}, month = {06/2013}, pages = {1569-1578}, author = {Sotirios P. Chatzis and D. Kosmopoulos and P. Doliotis} } @proceedings {90, title = {Developing visual competencies for socially assistive robots: the HOBBIT approach}, journal = {6th International Conference on Pervasive Technologies for Assistive Environments}, year = {2013}, abstract = {In this paper, we present our approach towards developing visual competencies for socially assistive robots within the framework of the HOBBIT project. We show how we integrated several vision modules using a layered architectural scheme. Our goal is to endow the mobile robot with visual perception capabilities so that it can interact with the users. We present the key modules of independent motion detection, object detection, body localization, person tracking, head pose estimation and action recognition and we explain how they serve the goal of natural integration of robots in social environments.}, author = {Kostas Papoutsakis and Pashalis Panteleris and Antonis Ntelidakis and Stavros Stefanou and Xenofon Zabulis and Dimitrios Kosmopoulos and Antonis Argyros} } @proceedings {89, title = {Robust offline topological map estimation using visual loop closures}, journal = {6th International Conference on Pervasive Technoogies for Assistive Environments}, year = {2013}, author = {Dimitrios Kosmopoulos and Ilias Maglogiannis and Fillia Makedon} } @article {11, title = {A system for multi-camera task recognition and summarization for structured environments}, journal = {Industrial Informatics, IEEE Transactions on}, volume = {9}, year = {2013}, pages = {161-171}, issn = {1551-3203}, doi = {10.1109/TII.2012.2212712}, author = {D. Kosmopoulos and A. Voulodimos and Doulamis, A.D.} } @article {2, title = {Bayesian Filter based Behavior Recognition in Workflows allowing for User Feedback}, journal = {Computer Vision and Image Understanding}, volume = {116}, year = {2012}, pages = {422-434}, abstract = {In this paper, we propose a novel online framework for behavior understanding, in visual workflows, capable of achieving high recognition rates in real-time. To effect online recognition, we propose a methodology that employs a Bayesian filter supported by hidden Markov models. We also introduce a novel re-adjustment framework of behavior recognition and classification by incorporating the user{\textquoteright}s feedback into the learning process through two proposed schemes: a plain non-linear one and a more sophisticated recursive one. The proposed approach aims at dynamically correcting erroneous classification results to enhance the behavior modeling and therefore the overall classification rates. The performance is thoroughly evaluated under real-life complex visual behavior understanding scenarios in an industrial plant. The obtained results are compared and discussed. }, author = {D. Kosmopoulos and N. Doulamis and A. Voulodimos} } @conference {86, title = {Hand Shape and 3D Pose Estimation Using Depth Data from a Single Cluttered Frame.}, booktitle = {International Symposium on Visual Computing (ISVC)}, year = {2012}, publisher = {Springer}, organization = {Springer}, keywords = {dblp}, isbn = {978-3-642-33178-7}, author = {Doliotis, Paul and Athitsos, Vassilis and Kosmopoulos, Dimitrios I. and Perantonis, Stavros J.}, editor = {Bebis, George and Boyle, Richard and Parvin, Bahram and Koracin, Darko and Fowlkes, Charless and Wang, Sen and Choi, Min-Hyung and Mantler, Stephan and Schulze, J{\"u}rgen P. and Acevedo, Daniel and Mueller, Klaus and Papka, Michael E.} } @article {40, title = {Improving Multi-Camera Activity Recognition by Employing Neural Network Based Readjustment}, journal = {Applied Artificial Intelligence}, volume = {26}, year = {2012}, pages = {97-118}, doi = {10.1080/08839514.2012.629540}, author = {Voulodimos, Athanasios S. and Doulamis, Nikolaos D. and Kosmopoulos, Dimitrios I. and Varvarigou, Theodora A.} } @proceedings {84, title = {A method for online analysis of structured processes using bayesian filters and echo state networks}, journal = {Proceedings of the 12th European Conference on Computer Vision - Volume Part III}, year = {2012}, pages = {71-80}, publisher = {Springer-Verlag}, address = {Florence - Italy}, isbn = {978-3-642-33884-7}, author = {Kosmopoulos, Dimitrios I. and Makedon, Fillia} } @article {21, title = {A Threefold Dataset for Activity and Workflow Recognition in Complex Industrial Environments}, journal = {MultiMedia, IEEE}, volume = {19}, year = {2012}, month = {July-Sept.}, pages = {42 -52}, issn = {1070-986X}, doi = {10.1109/MMUL.2012.31}, author = {Athanasios Voulodimos and D. Kosmopoulos and Vasileiou, Georgios and Sardis, Emmanuel and Anagnostopoulos, Vasileios and Lalos, Constantinos and Doulamis, A.D. and Varvarigou, T.A.} } @article {14, title = {Visual Workflow Recognition Using a Variational Bayesian Treatment of Multistream Fused Hidden Markov Models}, journal = {Circuits and Systems for Video Technology, IEEE Transactions on}, volume = {22}, year = {2012}, month = {July}, pages = {1076 -1086}, keywords = {active learning-based visual workflow recognition, Bayes methods, classification variance, data labeling, hidden Markov models, image sequences, learned model, learning (artificial intelligence), maximum a posteriori training, maximum information gain method, maximum likelihood training, MFHMM parameters, model variance, multistream fused hidden Markov models, overfitting issues, point estimate-based methods, posterior distribution, training methods, unlabeled data, variational Bayesian treatment, video streaming, video surveillance}, issn = {1051-8215}, doi = {10.1109/TCSVT.2012.2189795}, author = {Sotirios P. Chatzis and D. Kosmopoulos} } @proceedings {10, title = {A Dataset for Workflow Recognition in Industrial Scenes}, journal = {IEEE International Conference on Image Processing}, year = {2011}, pages = {3310-3313}, author = {A. Voulodimos and D. Kosmopoulos and Vasileiou, G. and Sardis, E. and Doulamis, A.D. and Anagnostopoulos, V. and Lalos, C. and Varvarigou, T.A.} } @article {31, title = {A hierarchical feature fusion framework for adaptive visual tracking}, journal = {Image and Vision Computing}, volume = {29}, year = {2011}, pages = {594 - 606}, keywords = {Particle filter, Sequential Monte-Carlo, Visual tracking}, issn = {0262-8856}, doi = {10.1016/j.imavis.2011.07.001}, author = {Alexandros Makris and Dimitrios Kosmopoulos and Stavros Perantonis and Sergios Theodoridis} } @article {26, title = {Multimodal and ontology-based fusion approaches of audio and visual processing for violence detection in movies}, journal = {Expert Systems with Applications}, volume = {38}, year = {2011}, pages = {14102 - 14116}, keywords = {Knowledge representation, Learning, Movie, Multimodal fusion, Ontology, Reasoning, Violence}, issn = {0957-4174}, doi = {10.1016/j.eswa.2011.04.219}, author = {Thanassis Perperis and Theodoros Giannakopoulos and Alexandros Makris and D. Kosmopoulos and Sofia Tsekeridou and Stavros J. Perantonis and Sergios Theodoridis} } @article {39, title = {Multiview Behavior Monitoring for Assistive Environments}, journal = {Universal Access in the Information Society}, volume = {10}, year = {2011}, pages = {115-123}, author = {D. Kosmopoulos} } @article {9, title = {Online classification of visual tasks for industrial workflow monitoring}, journal = {Neural Networks}, volume = {24}, year = {2011}, pages = {852{\textendash}860}, keywords = {Activity recognition, ESN, Fusion, Genetic algorithm, HMM, Workflow}, issn = {0893-6080}, author = {Athanasios Voulodimos and D. Kosmopoulos and Veres, Galina and Helmut Grabner and Van Gool, Luc and Varvarigou, T.A.} } @article {32, title = {Robust Jacobian matrix estimation for image-based visual servoing}, journal = {Robotics and Computer-Integrated Manufacturing}, volume = {27}, year = {2011}, pages = {82 - 87}, keywords = {Jacobian matrix, Robot manipulator control, Robust estimation, Semi-structured environments, Visual servoing}, issn = {0736-5845}, doi = {10.1016/j.rcim.2010.06.013}, author = {D.I. Kosmopoulos} } @article {24, title = {A variational Bayesian methodology for hidden Markov models utilizing Student{\textquoteright}s-t mixtures}, journal = {Pattern Recognition}, volume = {44}, year = {2011}, pages = {295 - 306}, keywords = {Violence detection}, issn = {0031-3203}, doi = {10.1016/j.patcog.2010.09.001}, author = {Sotirios P. Chatzis and D. Kosmopoulos} } @proceedings {18, title = {Video summarization guiding evaluative rectification for industrial activity recognition}, journal = {Computer Vision Workshops (ICCV Workshops), 2011 IEEE International Conference on}, year = {2011}, month = {nov.}, pages = {950-957}, keywords = {activity recognition framework, evaluative rectification method, feature extraction, image recognition, industrial activity recognition, industrial surveillance video, key-frame extraction, neural nets, neural network, production engineering computing, video signal processing, video summarization}, doi = {10.1109/ICCVW.2011.6130354}, author = {Voulodimos, A.S. and Doulamis, A.D. and D. Kosmopoulos and Varvarigou, T.A.} } @conference {101, title = {Audio-Visual Fusion for Detecting Violent Scenes in Videos}, booktitle = {Artificial Intelligence: Theories, Models and Applications, 6th Hellenic Conference on AI, SETN 2010, Athens, Greece, May 4-7, 2010. Proceedings}, year = {2010}, publisher = {Springer}, organization = {Springer}, author = {Theodoros Giannakopoulos and Alexandros Makris and Dimitrios I. Kosmopoulos and Stavros J. Perantonis and Sergios Theodoridis} } @proceedings {7, title = {Enhanced Human Behavior Recognition using HMM and Evaluative Rectification}, journal = {ACM Multimedia, ARTEMIS Workshop}, year = {2010}, author = {Doulamis, N.D. and Voulodimos, A.S. and D. Kosmopoulos and Varvarigou, T.A.} } @article {30, title = {Multiclass defect detection and classification in weld radiographic images using geometric and texture features}, journal = {Expert Systems with Applications}, volume = {37}, year = {2010}, pages = {7606 - 7614}, keywords = {Classification, Defects, Geometrical features, Radiography, Segmentation, Texture, Welds}, issn = {0957-4174}, doi = {10.1016/j.eswa.2010.04.082}, author = {Ioannis Valavanis and Dimitrios Kosmopoulos} } @proceedings {6, title = {Robust human behavior modeling from multiple cameras}, journal = {Pattern Recognition (ICPR), 20th International Conference on}, year = {2010}, pages = {3575-3578}, author = {D. Kosmopoulos and Voulodimos, A.S. and Varvarigou, T.A.} } @article {5, title = {Robust Visual Behavior Recognition}, journal = {Signal Processing Magazine, IEEE}, volume = {27}, year = {2010}, month = {Sep.}, pages = {34-45}, issn = {1053-5888}, author = {D. Kosmopoulos and Sotirios P. Chatzis} } @proceedings {8, title = {Robust Workflow Recognition Using Holistic Features and Outlier-Tolerant Fused Hidden Markov Models}, journal = {International Conference on Artificial Neural Networks}, volume = {1}, year = {2010}, pages = {551-560}, author = {Athanasios Voulodimos and Helmut Grabner and D. Kosmopoulos and Van Gool, Luc and Varvarigou, T.A.} } @article {28, title = {Detecting abnormal human behaviour using multiple cameras}, journal = {Signal Processing}, volume = {89}, year = {2009}, pages = {1723 - 1738}, keywords = {Behaviour understanding, hidden Markov model, Homography, Support vector machine, Trajectory}, issn = {0165-1684}, doi = {10.1016/j.sigpro.2009.03.016}, author = {Panagiota Antonakaki and Dimitrios Kosmopoulos and Stavros J. Perantonis} } @proceedings {73, title = {Robust Occlusion Handling with Multiple Cameras using a Homography Constraint}, journal = {Fourth International Conference on Computer Vision Theory and Applications}, volume = {2}, year = {2009}, pages = {560-565}, publisher = {INSTICC Press}, address = {Lisboa, Portugal}, author = {Anastasios L. Kesidis and Dimitrios I. Kosmopoulos} } @article {3, title = {Robust Sequential Data Modeling Using an Outlier Tolerant Hidden Markov Model}, journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence}, volume = {31}, year = {2009}, pages = {1657-1669}, issn = {0162-8828}, author = {Sotirios P. Chatzis and D. Kosmopoulos and Varvarigou, T.A.} } @article {27, title = {Vision-based production of personalized video}, journal = {Signal Processing: Image Communication}, volume = {24}, year = {2009}, pages = {158 - 176}, keywords = {Automated content production, Human identification, tracking}, issn = {0923-5965}, doi = {10.1016/j.image.2008.12.010}, author = {D.I. Kosmopoulos and A. Doulamis and Makris, A. and N. Doulamis and Sotirios P. Chatzis and S.E. Middleton} } @proceedings {15, title = {A robust approach towards sequential data modeling and its application in automatic gesture recognition}, journal = {Acoustics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on}, year = {2008}, month = {April}, pages = {1937-1940}, keywords = {automatic gesture recognition, data representation, finite Gaussian mixture model, Gaussian distribution, gesture recognition, hidden Markov model, hidden Markov models, hidden state distribution, maximum likelihood estimation, maximum likelihood framework, parameter estimation, sequential data modeling, student-t mixture model, time series, time series modeling}, doi = {10.1109/ICASSP.2008.4518015}, author = {Sotirios P. Chatzis and D. Kosmopoulos and Varvarigou, T.A.} } @article {12, title = {Signal Modeling and Classification Using a Robust Latent Space Model Based on t Distributions}, journal = {Signal Processing, IEEE Transactions on}, volume = {56}, year = {2008}, month = {March}, pages = {949 -963}, keywords = {Bayes methods, Bayesian inference, factor analysis modeling, inference mechanisms, robust latent space model, signal classification, signal modeling, signal processing, statistical distributions, t distribution, variational inference algorithm}, issn = {1053-587X}, doi = {10.1109/TSP.2007.907912}, author = {Sotirios P. Chatzis and D. Kosmopoulos and Varvarigou, T.A.} } @article {22, title = {Superquadric Segmentation in Range Images via Fusion of Region and Boundary Information}, journal = {Pattern Analysis and Machine Intelligence, IEEE Transactions on}, volume = {30}, year = {2008}, month = {may}, pages = {781 -795}, keywords = {Algorithms, Artificial Intelligence, Automated, computer vision, Computer-Assisted, edge detection, game theory, game-theoretic framework, Image Enhancement, Image Interpretation, image segmentation, Imaging, iterative methods, model-based edge detection, multiple object segmentation, Pattern Recognition, Reproducibility of Results, robot vision, Sensitivity and Specificity, Subtraction Technique, superquadric segmentation, Three-Dimensional}, issn = {0162-8828}, doi = {10.1109/TPAMI.2007.70736}, author = {Katsoulas, D.K. and Bastidas, C.C. and D. Kosmopoulos} } @article {17, title = {Automated Pressure Ulcer Lesion Diagnosis for Telemedicine Systems}, journal = {Engineering in Medicine and Biology Magazine, IEEE}, volume = {26}, year = {2007}, month = {Sept.-Oct.}, pages = {18 -22}, keywords = {automated pressure ulcer lesion diagnosis, biomedical imaging, diagnostic tools, digital images, image classification, medical diagnostic computing, region classification, remote patient location, telemedicine, telemedicine systems}, issn = {0739-5175}, doi = {10.1109/EMB.2007.901786}, author = {D. Kosmopoulos and Tzevelekou, F.L.} } @proceedings {4, title = {Hierarchical Feature Fusion for Visual Tracking}, journal = {IEEE Int Conf on Image Processing}, year = {2007}, pages = {289-292}, author = {Makris, A. and D. Kosmopoulos and Perantonis, S.J. and Theodoridis, S.} } @article {36, title = {Artificial Intelligence Systems for the Characterization of Digital Images Containing Pigmented Skin Lesions: Applications in the Detection of Malignant Melanoma}, journal = {Oncology Reports}, volume = {15}, year = {2006}, pages = {1027-1032}, author = {Maglogiannis, Ilias and D. Kosmopoulos} } @proceedings {23, title = {Box-like Superquadric Recovery in Range Images by Fusing Region and Boundary Information}, journal = {Pattern Recognition, 18th International Conference on}, volume = {1}, year = {2006}, pages = {719-722}, keywords = {box-like superquadric recovery, geometry, image reconstruction, image segmentation, object detection, piled box-like object localization, range imagery, recover-and-select framework, robotic bin-picking problem, superquadric segmentation}, doi = {10.1109/ICPR.2006.341}, author = {Katsoulas, D.K. and D. Kosmopoulos} } @proceedings {13, title = {Content-Based Time Sampling for Efficient Video Delivery over Networks of Low and Variable Bandwidth}, journal = {Digital Telecommunications, , 2006. ICDT {\textquoteright}06. International Conference on}, year = {2006}, month = {Aug.}, keywords = {bandwidth allocation, computational complexity, content management, content-based time sampling, data compression, decoding, image sampling, image sequences, linear frame skipping method, low bandwidth, MPEG compressed domain, multimedia communication, multimedia information, objective evaluation scheme, variable bandwidth, video coding, video communication, video delivery, video sequences}, doi = {10.1109/ICDT.2006.26}, author = {Doulamis, A.D. and D. Kosmopoulos and Doulamis, N.D.} } @article {37, title = {Hand Tracking for Gesture Recognition Tasks using Dynamic Bayesian Network}, journal = {International Journal of Intelligent Systems and Applications}, volume = {1}, year = {2006}, pages = {359-375}, author = {D. Kosmopoulos and Maglogiannis, Ilias} } @conference {85, title = {Violence content classification using audio features}, booktitle = {Proceedings of the 4th Helenic conference on Advances in Artificial Intelligence}, year = {2006}, publisher = {Springer-Verlag}, organization = {Springer-Verlag}, address = {Berlin, Heidelberg}, isbn = {3-540-34117-X, 978-3-540-34117-8}, author = {Giannakopoulos, Theodoros and Kosmopoulos, Dimitrios and Aristidou, Andreas and Theodoridis, Sergios} } @proceedings {20, title = {Content-based decomposition of gesture videos}, journal = {Signal Processing Systems Design and Implementation, 2005. IEEE Workshop on}, year = {2005}, month = {Nov.}, pages = {319-324}, keywords = {adapted video summary, binary trees, content-based decomposition, depicted gestures, feature extraction, gesture video decomposition, hierarchically structured video, human computer interaction, human-computer interface applications, key-frame extraction, video browsing, video signal processing, video transmission}, doi = {10.1109/SIPS.2005.1579886}, author = {Doulamis, N.D. and Doulamis, A.D. and D. Kosmopoulos} } @proceedings {19, title = {Gesture-based video summarization}, journal = {Image Processing, 2005. ICIP 2005. IEEE International Conference on}, volume = {3}, year = {2005}, month = {Sept.}, pages = {1220-1223}, keywords = {gesture energy, gesture-based video summarization, gestures extraction, image colour analysis, image segmentation, key-frames extraction, sign language videos, skin color segmentation, visual events, Zernike moments}, doi = {10.1109/ICIP.2005.1530618}, author = {D. Kosmopoulos and Doulamis, A.D. and N. Doulamis} } @article {35, title = {A system for the acquisition of reproducible digital skin lesions images}, journal = {Technology and Health Care}, volume = {11}, year = {2004}, pages = {425{\textendash}441}, keywords = {camera calibration, color constancy, image analysis, reproducible images, skin lesion inspection, telemedicine}, issn = {0928-7329}, author = {Maglogiannis, Ilias and Kosmopoulos, Dimitrios I.} } @article {25, title = {MD-SIR: a methodology for developing sensor{\textemdash}guided industry robots}, journal = {Robotics and Computer-Integrated Manufacturing}, volume = {18}, year = {2002}, pages = {403 - 419}, keywords = {Application development, Control objects library, Open robots controller architecture, Sensor integration}, issn = {0736-5845}, doi = {10.1016/S0736-5845(02)00031-5}, author = {D. Kosmopoulos and Varvarigou, T.A. and D.M Emiris and A.A Kostas} } @article {29, title = {Automated inspection of gaps on the automobile production line through stereo vision and specular reflection}, journal = {Computers in Industry}, volume = {46}, year = {2001}, pages = {49 - 63}, keywords = {Automated visual inspection, Gap measurement, Specular reflection}, issn = {0166-3615}, doi = {10.1016/S0166-3615(01)00113-0}, author = {Dimitrios Kosmopoulos and Theodora Varvarigou} } @proceedings {16, title = {An efficient depalletizing system based on 2D range imagery}, journal = {Robotics and Automation, 2001. Proceedings 2001 ICRA. IEEE International Conference on}, volume = {1}, year = {2001}, pages = {305-312}, keywords = {2D range imagery, active vision, depalletizing system, industrial robot, industrial robots, laser beam applications, materials handling, optical tracking, real time systems, real-time systems, time of flight laser sensor, tracking}, doi = {10.1109/ROBOT.2001.932570}, author = {Katsoulas, D.K. and D. Kosmopoulos} }