The power transmission lines are the link between power plants and the points of consumption, through substations. Most importantly, the assessment of damaged aerial power lines and rusted conductors is of extreme importance for public safety; hence, power lines and associated components must be periodically inspected to ensure a continuous supply and to identify any fault and defect. To achieve these objectives, recently, Unmanned Aerial Vehicles (UAVs) have been widely used; in fact, they provide a safe way to bring sensors close to the power transmission lines and their associated components without halting the equipment during the inspection, and reducing operational cost and risk. In this work, a drone, equipped with multi-modal sensors, captures images in the visible and infrared domain and transmits them to the ground station. We used state-of-the-art computer vision methods to highlight expected faults (i.e., hot spots) or damaged components of the electrical infrastructure (i.e., damaged insulators). Infrared imaging, which is invariant to large scale and illumination changes in the real operating environment, supported the identification of faults in power transmission lines; while a neural network is adapted and trained to detect and classify insulators from an optical video stream. We demonstrate our approach on data captured by a drone in Parma, Italy.

Owing to the development and widespread availability of cross-sectional imaging, in the last several decades, radiology has become pivotal in the diagnosis and management of many diseases. The use of medical imaging, including ionizing radiation-based modalities, continues to increase, raising concerns about patients’ radiation exposure,^1-4 with reported cumulative effective doses exceeding 100 mSv for single procedures.^5-8 Although accounting for only 17% of all medical examinations, multidetector computed tomography (CT) alone makes up approximately 50% of the total radiation burden for medical purposes, and a large multicenter trial by Rehani et al⁹ revealed that more than 1% of patients undergoing multiple CT examinations over 1 to 5 years received a cumulative effective dose above 100 mSv.

A person's well-being status is reflected by their face through a combination of facial expressions and physical signs. The SEMEOTICONS project translates the semeiotic code of the human face into measurements and computational descriptors that are automatically extracted from images, videos, and three-dimensional scans of the face. SEMEOTICONS developed a multisensory platform in the form of a smart mirror to identify signs related to cardio-metabolic risk. The aim was to enable users to self-monitor their well-being status over time and guide them to improve their lifestyle. Significant scientific and technological challenges have been addressed to build the multisensory mirror, from touchless data acquisition, to real-time processing and integration of multimodal data.

In recent times, with the exception of sporadic cases, the trend in computer vision is to achieve minor improvements compared to considerable increases in complexity. To reverse this trend, we propose a novel method to boost image classification performances without increasing complexity. To this end, we revisited ensembling, a powerful approach, often not used properly due to its more complex nature and the training time, so as to make it feasible through a specific design choice. First, we trained two EfficientNet‐b0 end‐to‐end models (known to be the architecture with the best overall accuracy/complexity trade‐off for image classification) on disjoint subsets of data (i.e., bagging). Then, we made an efficient adaptive ensemble by performing fine‐tuning of a trainable combination layer. In this way, we were able to outperform the state‐of‐the‐art by an average of 0.5% on the accuracy, with restrained complexity both in terms of the number of parameters (by 5–60 times), and the FLoating point Operations Per Second FLOPS by 10–100 times on several major benchmark datasets.