Despite the topic is outdated, this page is still in the search results. That's why it'll be useful to help people searching for pattern recognition. The growing computing power has opened new opportunities. Thanks to this, such tasks are now solving with the help of AI. Object recognition algorithms turned out to be too complicated, so now they are built using Machine Learning. The system analyzes tons of pictures with required types of patterns marked and finds dependencies. The process is not traditional programming. People's role here is to gather and markup the set of data that will be used to train AI and to set up the environment. This environment can be based on one of the existing frameworks like YOLOv3 Darknet, InceptionV3, SSD, Faster R-CNN, or other. And now some words about accuracy and performance. The most powerful projects, when trained properly, can scan up to several dozens of images per second with just 1-2% of errors. For example, SpotVision parking slots recognition software works in real-time and has an accuracy of 95%. And this is not the end! During the next few years, the AI-based systems will become recognize much more complex patterns that are currently unable to be analyzed with both traditional methods and ML-based approaches.