Road tunnels represent critical sections of the transport infrastructure where the risk of traffic crashes and fire incidents is significantly increased due to confined space, specific geometry, and limited visibility and ventilation conditions. This paper presents an analysis of traffic crashes that occurred on the motorway network in the Republic of Serbia during the period 2020–2024, with a special focus on the Šarani and Lipak tunnels. Data from the national crash database and the internal records of the public enterprise “Roads of Serbia” were used to identify the most frequent crash types and related risk factors. Although crashes in tunnels account for less than 0.1% of all recorded crashes, they show a notably higher proportion of severe outcomes. The analysis includes factors such as lighting conditions, pavement surface, traffic signalization, and the functionality of safety equipment. Furthermore, the paper outlines the national methodology for conducting Road Safety Inspections (RSI) in tunnels, harmonized with the current PIARC recommendations and the Serbian Rulebook on tunnel safety requirements. The results highlight the need for systematic monitoring of tunnel safety equipment and the implementation of risk-management measures to enhance the safety of road users in closed traffic environments.

The paper provides some practical aspects of frequency multiplication applied to the synchronization of power grid instruments operating at unknown frequency or in a regime with a constant small frequency drifts. The circuit that realizes the multiplication is intended to replace the device's internal clock and to guarantee fixed pre-defined (programmable) number of samples per period of the input signal. This prevents spectrum leakage and other issues associated with spectrum calculations from time samples using common time-frequency transformations. Consequently, the architecture of the device implementing this solution can be simplified through elimination of digital circuits for various filtering and numerical corrections. Two possible architectures for the realization of this device are presented and the restrictions necessary to ensure adequate positioning of harmonics along the frequency axis are shown.

Annotated medical imaging data remain scarce, and labels are often weak and noisy: in mammography, an examination comprises several high-resolution views, yet the diagnostic outcome is recorded only at the breast level. Such problems are naturally cast as Multiple Instance Learning (MIL), where the model must infer instance-level structure from bag-level labels alone. Although contemporary foundation encoders supply strong general-purpose embeddings, augmenting MIL data in this representation space remains an open problem as established techniques act on one instance at a time and ignore the statistical dependencies binding a bag together. We address this with SetFlow, a generative model that learns the distribution of complete MIL bags directly in a frozen encoder’s embedding space. SetFlow couples flow-matching training with a Set Transformer-inspired backbone, making it invariant to instance ordering while modeling intra-bag relationships. Generation is conditioned jointly on class label and per-instance scale, yielding coherent, semantically faithful bags rather than isolated vectors. Evaluating on two large public mammography datasets and two encoders, we assess distributional fidelity, nearest-neighbor behavior, and downstream augmentation utility. We show that generated bags reproduce real-data statistics and improve classification in certain configuration, with performance gains varying on the amount of synthetic data. An architecture ablation confirms each design choice contributes to performance.

The paper provides some practical aspects of frequency multiplication applied to the synchronization of power grid instruments operating at unknown frequency or in a regime with a constant small frequency drifts. The circuit that realizes the multiplication is intended to replace the device's internal clock and to guarantee fixed pre-defined (programmable) number of samples per period of the input signal. This prevents spectrum leakage and other issues associated with spectrum calculations from time samples using common time-frequency transformations. Consequently, the architecture of the device implementing this solution can be simplified through elimination of digital circuits for various filtering and numerical corrections. Two possible architectures for the realization of this device are presented and the restrictions necessary to ensure adequate positioning of harmonics along the frequency axis are shown.