Pedestrian safety at crosswalks remains a critical concern as traffic accidents frequently result from drivers’ failure to yield, leading to severe injuries or fatalities. In response, various jurisdictions have enacted pedestrian priority laws to regulate driver behavior. Nevertheless, intersections lacking clear traffic signage and environments with limited visibility continue to present elevated risks. The scarcity and difficulty of collecting data under such complex conditions pose significant challenges to the development of accurate detection systems. This study proposes a CyclicGAN-based loop-learning framework, in which the learning process begins with a set of manually annotated images used to train an initial labeling model. This model is then applied to automatically annotate newly generated synthetic images, which are incorporated into the training dataset for subsequent rounds of model retraining and image generation. Through this iterative process, the model progressively refines its ability to simulate and recognize diverse contextual features, thereby enhancing detection performance under varying environmental conditions. The experimental results show that environmental variations—such as daytime, nighttime, and rainy conditions—substantially affect the model performance in terms of F1-score. Training with a balanced mix of real and synthetic images yields an F1-score comparable to that obtained using real data alone. These results suggest that CycleGAN-generated images can effectively augment limited datasets and enhance model generalization. The proposed system may be integrated with in-vehicle assistance platforms as a supportive tool for pedestrian-crossing detection in data-scarce environments, contributing to improved driver awareness and road safety.

Riding comfort and safety are essential requirements for any form of transportation but particularly for electric bicycles (e-bikes), which are highly affected by varying road conditions. These factors largely depend on the effectiveness of the e-bike’s control strategy. While several studies have proposed control approaches that address comfort and safety, vibration—an influential factor in both structural integrity and rider experience—has received limited attention during the design phase. Moreover, many commercially available e-bikes provide manual assistance-level settings, leaving comfort and safety management to the rider’s experience. This study proposes a Road-Adaptive Vibration Reduction System (RAVRS) that can be deployed on an e-bike rider’s smartphone to automatically maintain riding comfort and safety using manual assistance control. A fuzzy-based control algorithm is adopted to dynamically select the appropriate assistance level, aiming to minimize vibration while maintaining velocity and acceleration within thresholds associated with comfort and safety. This study presents a vibration analysis to highlight the significance of vibration control in improving electronic reliability, reducing mechanical fatigue, and enhancing user experience. A functional prototype of the RAVRS was implemented and evaluated using real-world data collected from experimental trips. The simulation results demonstrate that the proposed system achieves effective control of speed and acceleration, with success rates of 83.97% and 99.79%, respectively, outperforming existing control strategies. In addition, the proposed RAVRS significantly enhances the riding experience by improving both comfort and safety.

Computer-assisted learning can help erhu learners to analyze their playing performance and identify areas for improvement. Therefore, in this study, a computerized system based on a You Only Look Once (YOLO)-OD model was developed for erhu bowing training. First, Therblig analysis was performed to segment the erhu bowing process into multiple steps, which were then analyzed thoroughly to identify critical objects for detection. Second, a YOLO-OD model was developed to detect and track the critical objects identified in video frames. Third, scoring methodologies were developed for bow level and bow straightness. The YOLO-OD model and the aforementioned scoring methodologies were incorporated into a computerized training system for erhu bowing, which enables erhu learners to practice independently. It provides scores for bow level and bow straightness, allowing learners to evaluate their technique, as well as feedback and instant alerts regarding incorrect motions and postures, which enable learners to adjust their actions and postures in real time. In addition, teachers or coaches can refer to the videos and other data collected using the proposed system in order to identify problematic erhu bowing techniques and provide students with relevant suggestions and feedback.