© 2021 IEEE.Detection of traffic sign is quite important in autonomous driving and driver assistance systems. In this study, Faster R-CNN as a very powerful object detection system is integrated with a feature extractor Inception Resnet V2 system, and it is evaluated and tested for detection and recognition of traffic sign. In addition, to enhance the effectiveness of this presented architecture, a new procedure of preprocessing the image to augment the detection performance of a traffic sign detector is also proposed. This new image preprocessing is introduced to reduce noise by blurring statistically some regions in an image, in which existence possibility of a traffic sign is lower. Hence, a more efficient training of object detector approach is presented. In the experiments, the publicly available Faster R-CNN Inception Resnet V2 model pre-trained on Microsoft COCO dataset is imported from Tensorflow model zoo on Github and is trained again on German Traffic Sign Detection Benchmark dataset. Also, to recognize detected traffic signs, German Traffic Sign Recognition Benchmark is used to train a CNN classifier. The experimental study illustrates the effectiveness of the presented approach while benchmarking with existing methodologies.