Industrial quality control systems in mass production facilities must exhibit a very high level of defect detection efficiency. The continuous increase in quality control and process automation requirements is leading companies to increasingly experiment with artificial intelligence methods to boost efficiency. One potential application area for AI is visual inspection, which is an essential element of almost every quality control process. In this article, we propose the use of neural networks for the visual inspection of rotationally symmetric components. The presented method leverages the existence of symmetry to represent images in a polar coordinate system and to implement the learning process on data modified in this way. An undeniable advantage of the proposed algorithm is also the transition from a two- dimensional to a one-dimensional representation, which significantly reduces the demand for memory resources and the required computational power. This is particularly important in mass production processes, where the time for data analysis is relatively short. The high repeatability of images due to the mass production nature makes this model exceptionally effective, allowing not only to confirm the presence of defects but even to locate them. The obtained results are compared with the results achieved using a convolutional neural network operating on two-dimensional images.