Real-world applications play an important role in CILab research. We develop and evaluate Artificial Intelligence methods in domains where data are heterogeneous, operating conditions are imperfect and requirements such as efficiency, interpretability and reliability become as important as predictive performance.
Cultural heritage and digital humanities are among the laboratory’s most established application domains. Our research investigates computational methods for analysing visual artworks and cultural collections, combining computer vision and deep learning with structured knowledge, knowledge graphs and multimodal models. These activities range from artwork classification and retrieval to semantic analysis, generative approaches and the exploration of relationships between visual and contextual information.
In healthcare and biomedical applications, we investigate Computational Intelligence and deep learning methods for medical images, physiological signals and heterogeneous clinical data. Explainability plays a particularly important role in this domain, where model predictions need to be interpreted and critically assessed.
In drone-based visual intelligence, monitoring and emergency response, we investigate AI methods for autonomous and remotely operated aerial platforms. These scenarios combine computer vision with multimodal sensing and resource-efficient AI, while introducing practical constraints related to onboard computation, environmental conditions, real-time operation and system reliability. Applications include road safety, emergency response and intelligent monitoring.
Our research also extends to other complex domains, including sign language understanding and translation, agriculture and safety-critical systems, providing further settings in which methodological advances can be tested against concrete problems.
We do not regard these domains simply as isolated case studies. Real-world applications expose limitations that conventional benchmarks may hide and generate new research questions concerning limited data, heterogeneous information, computational efficiency, explainability and robustness. This interaction between methodological research and experimentation is a defining aspect of CILab.