Deep learning is an artificial intelligence (AI) function that specially based on Convolutional neural networks (CNNs) techniques that simulate the workings of human brain for processing data and are able to learn without human supervision. This techniques are widely used in many applications of pattern and image recognition with obtaining a number of advantages compared to other traditional techniques. This study covers the basics of CNNs including a description to the functions of its various layers and mathematically learns how to use it in many AI applications. The structure models of these applications that are based on CNNs are proposed to show the improvement in processing time and state-of-the-art precision. This study also provides an overview of different CNNs architectures that are used in famous useful applications for different fields in our life based on computer vision tasks.