With the development of image processing techniques and the growing GPU computing power, these techniques are very common in real-time systems such as self-driving cars. Photo-separation methods, which allow us to locate objects in the image and mark pixels of these objects, are image processing techniques commonly used in self-driving cars. In addition to the study of classical photographs, detailed reading-based techniques are widely available in textbooks. This paper discusses why photo classification methods are important for self-driving cars. We explore ways to classify images into two main groups: semantic classification and example. We see that these methods are used in many different situations in self-driving cars and how important they are in safe travel.