الفهرس 
Chapter 1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 87 |  |  |
| | | from | 87 | | |
Abstract
Recently, autonomous driving technology has been developed to self-driving safely and smoothly. However, convolutional neural networks (CNNs) have impacted autonomous driving technology. Finding the best CNNs designs for safe driving is a difficult task. Therefore, there is always a need for experts to create most of them. Traffic Sign Classification may help drivers to follow up the traffic laws and avoid accidents. Also, self-driving vehicles use sensors and cameras to detect objects and obstacles to make tracking and traffic monitoring easier in real-time.
This study uses genetic algorithms to construct automated genetic behavior cloning (AGBC). The “Automated” features of the proposed algorithm require little knowledge of CNNs. Thus, our system is a combination of the AGBC and CNN giving AGBC-CNN. Using a front-facing camera and experienced driver’s steering directions, researchers can build strong CNN architectures for learning safe driving behavior. Furthermore, it can be used to train our system AGBC-CNN to emulate the human driving behavior. We compared our results with four manually adjusted behavior cloning CNNs (MABC-CNNs) and one automated genetically/manually adjusted behavior cloning CNNs (AGMABC-CNNs) to validate its effectiveness and robustness. Our study shows that the AGBC-CNN outperforms existing architecture design techniques (MABC-CNNs and AGMABC-CNN) in terms of classification accuracy.
Also, to obtain the state-of-the-art performance, we used several transformations to the input images, including rotation, translation, shearing, and magnification. We were able to achieve a 99.46% accuracy rate on the German Traffic Sign Recognition Benchmark (GTSRB) dataset.
Also, we use the YOLO method, which stands for ”You Only Look Once,” to detect objects. Train our algorithm on 117k images, including items from 80 categories, and use a S X S grid to extract the most necessary details. That achieved high efficiency in the detection shown in visualization results.