MARINA identifies number-plates with an accuracy of 99.3% even when the plate is dirty, the viewing angle is large in different scenarios shot from a fixed viewpoint or on mobile devices. The module is based on efficient algorithms of image processing, cascaded schemes for recognition and also uses its own technology of ultra-light and extremely-precise neural networks.

The module recognizes the number-plates of Russia, Belarus, Ukraine, Japan, Mexico, China, Poland, Spain, France, Sweden and Finland. The number of countries supported in the system is constantly growing. MARINA can be used with small compact computing devices, such as droid xu4 and raspberry pi 3, and can be run on Windows XP, Vista, 7, 8, 10, Linux, iOS, Android, and x86, x64, ARM, Elbrus architectures.

#recognition#neural_networks#computer_vision#artificial_intelligence

#Visillect