FORTH Institute of Computer Science

Historical newspapers are invaluable resources that capture the cultural, social, economic, and political dimensions of their time, providing unique insights into past events and daily life. However, the potential of these archives remains underutilised due to the limitations of traditional research methods, which rely on laborious manual examination of individual pages or articles. Digital archives have begun to address this issue by enhancing accessibility, but the lack of effective tools for parsing and analysing the contents still restricts their utility for large-scale historical research. Optical Character Recognition (OCR) technology plays a foundational role in converting scanned images into searchable text. However, OCR alone cannot adequately handle the complexities of historical newspapers, which often feature degraded image quality, inconsistent typography, and intricate layouts. In parallel, the effective segmentation and ordering of newspaper contents is essential for generating well-structured and logically ordered data from specific newspaper sections. This paper focuses on the development of a pipeline that addresses these challenges, using the daily newspaper Le Sémaphore de Marseille as a case study (35,703 issues covering the period 1827 to 1944). The proposed system integrates layout analysis, OCR, and Generative AI-driven information extraction techniques to extract specific data elements from scanned images, namely ship arrivals data, and to store them in a machine-readable format for further analysis. This well-structured output enables historians to analyse historical trade patterns, economic trends, and regional interactions more efficiently and at a much larger scale. An evaluation of the key pipeline components demonstrates the effectiveness of the system, achieving an F1-score of 96% in paragraph segmentation and information extraction, while also highlighting the need for further improvement in the critical OCR component.

Abstract State-of-the-art computational models of vision largely focus on fitting trial-averaged spike counts to visual stimuli using overparameterized neural networks. However, a computational model of the visual cortex should predict the dynamic responses of neurons in single trials across different experimental conditions. In this study, we investigated trial-by-trial inter-areal interactions in the visual cortex by predicting neuronal activity in one area based on activity in another, distinguishing between stimulus-driven and non-stimulus-driven shared variability. We analyzed two datasets: calcium imaging from mouse V1 layers 2/3 and 4, and extracellular neurophysiological recordings from macaque V1 and V4. Our results show that neuronal activity can be predicted bidirectionally between L2/3 and L4 in mice, and between V1 and V4 in macaques, with the latter interaction exhibiting directional asymmetry. The predictability of neuronal responses varied with the type of visual stimulus, yet responses could also be predicted in the absence of visual stimulation. In mice, we observed a bimodal distribution of neurons, with some neurons primarily driven by visual inputs and others showing predictable activity during spontaneous activity despite lacking consistent visually evoked responses. Predictability also depended on intrinsic neuronal properties, receptive field overlap, and the relative timing of activity across areas. Our findings highlight the presence of both stimulus- and non-stimulus-related components in interactions between visual areas across diverse contexts and underscore the importance of non-visual shared variability between visual regions in both mice and macaques.

The two terms of our title have long been destined for each other but, in the style of a Hollywood romantic comedy, have taken their sweet time to merge in an embrace. This special issue wants to be a setting for such a moment. We come into this with a double-pronged hypothesis: that “Beat time” is queer time, as it disrupts and rearranges the chronological patterns of straight society, and that recent critical work on queer temporality—the particular rhythms and tempos of queer minorities—c...

Cross-border collaboration among Law Enforcement Agencies is essential for effective and timely suspect identification, especially when the availability of biometric data varies between agencies. This paper presents a scalable and secure approach for multimodal biometric identification across multiple jurisdictions. Our approach allows Law Enforcement Agencies to combine biometric modalities -facial images, fingerprints, and voice samples- and compare them with collaborating agencies, improving the overall accuracy and effectiveness of suspect identification. By leveraging deep learning models for indexing and comparison, efficient data retrieval was achieved without compromising privacy or security. To ensure the protection of sensitive biometric data, our approach incorporates advanced encryption mechanisms, including Homomorphic Encryption for secure computations and Advanced Encryption Standard (AES encryption) for safeguarding biometric information. Its decentralised architecture allows each Law Enforcement Agency to maintain independent instances of the Deep Learning Indexer and Comparator, minimising risks associated with centralising sensitive data and supporting seamless collaboration between agencies. This approach not only improves the accuracy of suspect identification but also enhances operational efficiency by allowing Law Enforcement Agencies to query and share biometric data securely across borders.

Soil moisture is an essential climate variable that directly influences many hydrological, agricultural, and water-cycle processes. Many satellites have been launched and are still being launched to map soil moisture accurately at a global scale. Motivated by the coarse resolution of existing satellite products, many statistical, physics-based, and machine learning-based methods have been proposed to downscale soil moisture to much finer spatial scales. In this paper, we propose a novel deep learning approach that is constrained by the Tau-omega radiative transfer model to enhance the resolution of surface soil moisture. We demonstrate the proposed framework by downscaling Soil Moisture Active Passive (SMAP) L-band brightness temperature (TB) with C-band synthetic aperture radar (SAR) backscattering coefficient (σ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</inf>) imagery from Sentinel-1A/B and subsequently retrieving high spatial resolution (1km) soil moisture.

Unhealthy behavior increases the risk of dementia. Various socio-cognitive determinants influence whether individuals persist in or alter these unhealthy behaviors. This study identifies relevant determinants of behavior associated to dementia risk. 4,104 Dutch individuals (40-79 years) completed a screening questionnaire exploring lifestyle behaviors associated with dementia risk. Subsequently, 3,065 respondents who engaged in one or more unhealthy behaviors completed a follow-up questionnaire investigating socio-cognitive determinants of these behaviors. Cross-tables were used to assess the accuracy of participants' perceptions regarding their behavior compared to recommendations. Confidence Interval-Based Estimation of Relevance (CIBER) was used to identify the most relevant determinants of behavior based on visual inspection and interpretation. Among the respondents, 91.3% reported at least one, while 65% reported two or more unhealthy lifestyle behaviors associated to dementia risk. Many of them were not aware they did not adhere to lifestyle recommendations. The most relevant determinants identified include attitudes (i.e., lacking a passion for cooking and finding pleasure in drinking alcohol or smoking), misperceptions on social comparisons (i.e., overestimating healthy diet intake and underestimating alcohol intake), and low perceived behavioral control (i.e., regarding changing physical inactivity, altering diet patterns, and smoking cessation). Individual-level interventions that encourage lifestyle change should focus on enhancing accurate perceptions of behaviors compared to recommendations, while strengthening perceived control towards behavior change. Given the high prevalence of dementia risk factors, combining interventions at both individual and environmental levels are likely to be the most effective strategy to reduce dementia on a population scale.

This paper describes the motivation, framework, concepts, method, implementation, and results of the HCII2023 Design Café, a dedicated participatory highly interactive design workshop held in Copenhagen in July 2023. Motivated by the ubiquitous challenges our world is facing, this initiative had the goal to explore six main issues from an interdisciplinary perspective with a focus on the UN SDG 11: “Sustainable Cities and Communities” and the HCI Grand Challenge 2 “Human-Environment Interactions.” The six issues were formulated as questions and presented to the participants working in small groups. (1) How to create inclusive and ethical smart cities? (2) How to establish trust between people and smart environments? (3) How to address privacy concerns in smart environments that adopt the “disappearing computer” paradigm? (4) How to promote explainability and transparency of policies and measures to citizens of smart cities or in smart environments in general? (5) How to design incentives and rewards for engagement and sustainable behavior in smart cities at a personal as well as collective/corporate level? (6) How to measure success and impact in sustainable smart city projects? The method and approach of the Design Café is a tailored composition of a guided, structured format combined with and inspired by processes of informal communication and exchange of knowledge and ideas. Aligned moderation, a minimum set of rules, a set of relevant topics and an interdisciplinary group of motivated participants working in rotating formations provides the structure for achieving results. The results show that the six issues are not independent of each other, but require a holistic view, considering the various dependencies as well as synergies when exploring solutions. Nevertheless, the importance of first establishing higher-level goals based on ethical and inclusive approaches fostering human dignity and human rights were key. Acceptance of overall goals, processes, rules, and regulations was considered as the fundamental pre-requisite for sustainable change towards a declared goal. Acceptance needs trust and privacy as well as explainability and transparency of policies and measures. The role of incentives and rewards for engagement and sustainable behavior was twofold. Offering incentives must include planning on how to measure their impact. Effective measurement of success and impact depends heavily on how the institutions address privacy. Explainability and transparency should become one of the ethical guidelines that steer and control concepts, decision making, and implementations of all activities. The scale of these societal challenges still needs to be recognized by those responsible. It is essential to educate decision makers in the psychological needs and effects of sense making, comprehensibility, and finally acceptance and well-being moving towards a humanity-centered design.

With the increase of AI applications in the field of 3D estimation of hand state, the quality of the datasets used for training the relevant models is of utmost importance. Especially in the case of datasets consisting of real-world images, the quality of annotations, i.e., how accurately the provided ground truth reflects the true state of the scene, can greatly affect the performance of downstream applications. In this work, we propose a methodology with significant impact on improving ubiquitous 3D hand geometry datasets that contain real images with imperfect annotations. Our approach leverages multi-view imagery, temporal consistency, and a disentangled representation of hand shape, texture, and environment lighting. This allows to refine the hand geometry of existing datasets and also paves the way for texture extraction. Extensive experiments on synthetic and real-world data show that our method outperforms the current state of the art, resulting in more accurate and visually pleasing reconstructions of hand gestures.