This thesis describes the functionality and usefulness of AI vision as a process that simulates human vision and has a focus on neural networks. The realization itself is gradual, starting from the basic concepts of perceptrons to intricate layers of convolutional neural networks. Terms related to neural networks and their function within the network are defined in detail. Also, the structure of different neural networks, types of activation functions, advantages and disadvantages of their application in various situations are listed. The focus of the work is the training process and the analysis of the obtained results. The basic concept is based on the application of machine learning, which uses visual inputs such as images and videos to participate in the detection of objects. The importance of convolutional neural networks for artificial vision is highlighted. The YOLO (You Only Look Once) neural network developed with the aim of applying artificial intelligence for object detection was presented and used. Using the YOLO software, the project task was realized based on a set of input data, the training process and the code implementation of the object detection model. The interpretation of the results highlighted the need for the application of artificial intelligence in the detection of objects and, in particular, their role as an aid to the visually impaired.