The automatic detection of violent actions in public places through video analysis is difficult because the employed Artificial Intelligence-based techniques often suffer from generalization problems. Indeed, these algorithms hinge on large quantities of annotated data and usually experience a drastic drop in performance when used in scenarios never seen during the supervised learning phase. In this paper, we introduce and publicly release the Bus Violence benchmark, the first large-scale collection of video clips for violence detection on public transport, where some actors simulated violent actions inside a moving bus in changing conditions, such as the background or light. Moreover, we conduct a performance analysis of several state-of-the-art video violence detectors pre-trained with general violence detection databases on this newly established use case. The achieved moderate performances reveal the difficulties in generalizing from these popular methods, indicating the need to have this new collection of labeled data, beneficial for specializing them in this new scenario.

Tracking and action-recognition algorithms are currently widely used in video surveillance, monitoring urban activities and in many other areas. Their development highly relies on benchmarking scenarios, which enable reliable evaluations/improvements of their efficiencies. Presently, benchmarking methods for tracking and action-recognition algorithms rely on manual annotation of video databases, prone to human errors, limited in size and time-consuming. Here, using gained experiences, an alternative benchmarking solution is presented, which employs methods and tools obtained from the computer-game domain to create simulated video data with automatic annotations. Presented approach highly outperforms existing solutions in the size of the data and variety of annotations possible to create. With proposed system, a potential user can generate a sequence of random images involving different times of day, weather conditions, and scenes for use in tracking evaluation. In the design of the proposed tool, the concept of crowd simulation is used and developed. The system is validated by comparisons to existing methods.

The advancement of deep learning methods across various applications has forced the creation of enormous training datasets. However, obtaining suitable real-world datasets is often challenging for various reasons. Consequently, numerous studies have emerged focusing on the generation and utilization of synthetic data in the training process. Hence, there is no universal formula for preparing synthetic data and leveraging it in network training to maximize the effectiveness of various detection methods. This work provides a comprehensive overview of several synthetic data generation techniques, followed by a thorough investigation into the impact of training methods and the selection of synthetic data quantities. The outcomes of this research enable the formulation of conclusions regarding the recipe for developing synthetic data with high efficacy in enhancing detection methods. The main conclusion for the synthetic data generation methods is to ensure maximum diversity at a high level of photorealism, which allows improving the classification quality by more than 5% to even 19% for different detection metrics.

Purpose: The purpose of this work was to examine the suitability of VIPAR^nd polymer gel–9.4 T magnetic resonance microimaging system for high spatial resolution dose distribution measurements.Methods: The VIPAR^nd samples (3 cm in outside diameter and 12 cm in height) were exposed to ionizing radiation by using a linear accelerator (Varian TrueBeam, USA; 6 MV x‐ray beam). In the calibration stage, nine gel dosimeter vials were irradiated in a water phantom homogenously to the doses from 1.5 to 30 Gy in order to obtain R2‒dose relation. In the verification stage, two gel dosimeter vials were irradiated in the half beam penumbra area of 10 × 10 cm radiation field using the amount of monitor units appropriate to deliver 20 Gy at the field center. The gels were imaged on a vertical 9.4 T magnetic resonance (MR) microimaging scanner using single slice and multislice (9 slices) multiecho (90 × 7 ms) sequences at the spatial resolutions of 0.2–0.4 × 0.2–0.4 × 3 mm³ and 0.2–0.4 × 0.2–0.4 × 1 mm³ respectively. The gels were subjected to microimaging during the period of two weeks after irradiation. The reference data consisted of the dose profiles measured using the diode dosimetry, radiochromic film, ionization chamber, and the water phantom system.Results: The VIPAR^nd‒9.4 T MR microimaging system was characterized by the dose sensitivity of 0.067 ± 0.002 Gy⁻¹ s⁻¹ at day 3 after irradiation. The dose resolution at 10 Gy (at P = 95%) was equal to 0.42 Gy at day 3 after irradiation using a single slice sequence (0.2 × 0.2 × 3 mm³) and 2.0 Gy at day 4 after irradiation using a multislice sequence (0.2 × 0.2 × 1 mm³) for one signal acquisition (measurement time: 15 min). These values were improved by ~1.4‐fold when using four signal acquisitions in the single slice sequence, and by ~2.78‐fold for 12 signal acquisitions in the multislice sequence. Furthermore, decreasing the in‐plane resolution from 0.2 × 0.2 mm² to 0.4 × 0.4 mm² resulted in a dose resolution of 0.3 Gy and 1 Gy at 10 Gy (at P = 95%) for one signal acquisition in the single slice and multislice sequences respectively (measurement time: 7.5 min).As reveals from the gamma index analysis the dose distributions measured at days 3–4 postirradiation using both VIPAR^nd verification phantoms agree with the data obtained using a silicon diode, assuming 1 mm/5% criterion. A good interphantom reproducibility of the polymer gel dosimetry was proved by monitoring of two phantoms up to 10 days after irradiation. However, the agreement between the dose distributions measured using the diode and polymer gel started to get worse from day 5 after irradiation.Conclusion: The VIPAR^nd–9.4T MR microimaging system allows to obtain dose resolution of 0.42 Gy at 10 Gy (at P = 95%) for a spatial resolution of 0.2 × 0.2 × 3 mm³ (acquisition time: 15 min). Further studies are required to improve a temporal stability of the gel‐derived dose distribution.