TY - JOUR AU - Quoc-Viet Hoang, AU - Trung-Hieu Le, AU - Minh-Quy Nguyen, PY - 2020/10/12 Y2 - 2024/03/28 TI - RD: A NOVEL DATASET FOR OBJECT DETECTION IN RAINY WEATHER CONDITIONS JF - UTEHY Journal of Science and Technology JA - jst VL - 27 IS - SE - Articles DO - UR - http://tapchi.utehy.edu.vn/index.php/jst/article/view/394 SP - 86-90 AB - In recent years, object detection models using convolutional neural networks have received considerable attention and achieved appealing results in autonomous driving systems. However, in inclement weather conditions, the performance of these models decreases tremendously due to the lack of relevant datasets for training strategies. In this work, we address the problems of object detection with rain interference by introducing a novel rain driving dataset, named RD. Our dataset highlights a variety of data sources with 1,100 real-word rainy images depicting a variety of driving scenes and comes with ground truth bounding box annotations for five common traffic object categories. We adopt RD for training three state-of-the-art object detection models, encompassing SSD512, RetinaNet, and YOLO-V3. Experimental results show that the performance of SSD512, RetinaNet, and YOLO-V3 models are advanced up to 5.64%, 8.97%, and5.70%, respectively. ER -