Recognizing Bangladeshi Traffic Signs in the Wild

Traffic sign detection is an indispensable part of autonomous driving and transportation safety systems. However, accurate detection and recognition of traffic signs remain challenging, especially under extreme conditions, such as various weather and geo-social features. Though a lot of work has been done in the domain of Traffic Sign Detection and Recognition (TSDR), only a few of them focus on a dataset that comprises a wide variety of real-world challenges. Moreover, in the context of Bangladeshi traffic sign detection, the research is in a very preliminary stage, whereas, the geo-social features of Bangladesh add some unique challenges that are not seen in most parts of the world. In this regard, we have curated a dataset containing 2986 images belonging to 15 different classes of Bangladeshi traffic signs collected under conditions like varying distance, occlusion, blurry conditions, geological variations, varying lighting conditions, etc., reflecting several real-world scenarios. We have provided a thorough performance analysis with different state-of-the-art object detection algorithms where the YOLOv7 architecture has been found to be the best-performing model with a mAP value of 0.889, making it a suitable model for real-life applications.