Pedestrian and Objects Detection by Using Learning Complexity-Aware Cascades

Due to the considerable technological development in all joints of life, the trend has become significant towards automating various processes in daily life, such as smart cities, the Internet of things, and cloud services. One of the most crucial applications is self-driving cars, which will be a quantum leap in this field. The main problem with these vehicles will be how to provide the necessary accuracy to deal with various situations, such as sudden stops and pedestrian crossing. In this paper, we propose an effective method for automating autonomous vehicles by improving their ability to make appropriate decisions at the right time. For this, we rely on sequential training that is aware of the complexity. The system is trained and provided to the vehicles, where the presence of pedestrians is detected using machine learning algorithms, such as a deep convolutional neural network (CNN). The findings obtained in this research suggest a clear improvement in the vehicle's ability to make decisions and a great speed in responding to the event and parking the vehicle when passing by.