Abstract: Autonomous driving is one of the interesting research areas of modern times and traffic sign detection is a very important and crucial problem in this research. In this project, we propose to explore the YOLO Architecture and its compatibility to solve this problem. The objective is locating and classification of traffic signs in natural street scenes.The key challenge to be addressed in this problem is recognition of minute targets in an extended and complex image background. Other object detection models like Fast RCNN and Faster R-CNN[1] have been used for this problem. The main drawbacks with such methods is their speed - they fail to be real time. Thus, the motivation behind exploring YOLO for this task is the speed - it is about 6× faster than faster R-CNN. In this work, we also propose a novel modified loss function for the YOLO model to perform better for traffic sign detection.Keywords: Traffic sign recognition, CNN, Faster R-CNN,YOLOv5