ABSTRACT Traffic sign detection (TSD) plays a critical role in real‐world applications. Advanced TSD frameworks have achieved promising results on public TSD datasets. Identifying and detecting traffic signs in challenging environments has garnered significant attention, where the accurate recognition of obscured and small signs is a critical challenge. Recent frameworks have achieved remarkable performance but struggle to detect small, occluded signs and recognise them under diverse atmospheric conditions. To address these problems, we propose a multiscale feature fusion and multiscale context‐aware aggregation (MSCA) framework using You Only Look Once Version 9 (YOLOv9) for TSD. Multiscale feature fusion helps filter out less useful features, leading to more informative representations, and combines fine‐grained lower‐level features with semantically rich higher‐level features. In addition, MSCA fusion captures broader contexts without compromising spatial resolution and increases the receptive field without increasing parameters or computational cost, unlike methods that involve increasing kernel size or pooling. The MSCA‐YOLO method enhances the extracted features, focusing more on meaningful input areas, minimising the effects of size and weather differences, and improving the overall robustness of the system. The extensive experiments on the Tsinghua‐Tencent 100K and Changsha University of Science and Technology Chinese Traffic Sign Detection Benchmark 2021 datasets thoroughly assess the proposed method, achieving state‐of‐the‐art results.
Mahadshetti et al. (Wed,) studied this question.