We present a deep learning framework leveraging the ResUNet-a framework for pixel-wise semantic segmentation of cracks and potholes. By integrating key components including a U-Net encoder/decoder backbone, residual connections, atrous convolutions, pyramid scene parsing pooling, and multi-tasking inference, the proposed method exhibits robustness in capturing intricate spatial details and inter-pixel contextual relationships essential for accurate road defect segmentation. Experimental results validate the efficacy of the proposed approach, with ResUNet-a consistently surpassing the conventional U-Net model, demonstrating its superior performance in crack and pothole segmentation tasks and thus providing a useful auxiliary tool for road maintenance and safety.

Our proposed framework aims to efficiently train a network of fellows on the field of the resilience of Cultural Heritage (CH) areas and historic cities against Climate Change (CC) and other types of hazards. Towards this direction, the proposed framework aims to introduce a research framework for downscaling the created climate and atmospheric composition as well as associated risk maps down to the 1x1 km (historic area) scale, and specific damage functions for CH ma-terials. Applying atmospheric modelling for specific CC scenarios at such refined spatial and time scales allows for an accurate quantitative and qualitative impact as-sessment of the estimated micro-climatic and atmospheric stressors. Our proposed framework will perform combined structural/geotechnical analysis of the CH sites and damage assessment under normal and changed conditions, based on the cli-matic zone, the micro-climate conditions, the petrographic and textural features of building materials, historic data for the structures, the effect of previous restoration processes and the environmental/physical characteristics of the surrounding envi-ronment. The data coming from installed monitoring system will be coupled with simulated data and will be further analysed through our data management system, while supporting communities’ participation and public awareness. The data from the monitoring system will feed the Decision- Support-System (DSS) so as to pro-vide proper adaptation and mitigation strategies. The produced vulnerability map will be used by the local authorities to assess the threats of CC (and other natural hazards), visualize the built heritage and cultural landscape under future climate scenarios, model the effects of different adaptation strategies, and ultimately prior-itize any rehabilitation actions to best allocate funds in both pre- and post-event en-vironments. To train the fellows, our approach will make use of extensive workshop and training sessions, as well organise summer schools.

In recent years, there has been a notable rise in the adoption of Unmanned Aerial Vehicles (UAVs) for infrastructure inspection. This study focuses on presenting IKAROS, an integrated system that harnesses the combined capabilities of multiple UAVs and deep learning (DL) tools to improve the efficiency of road transport network inspections, facilitate the detection of road defects, and manage traffic emergencies. Specifically, we introduce the IKAROS system, outlining its key components and functionalities, with a particular emphasis on the applied technologies and the interconnections of its various systems. In addition, we provide an overview of the pilot site and address the challenges associated with real-world implementation.