Comparative analysis of identification of dynamic objects by scale-invariant feature transform and deep neural networks
Evgeniy Ivliev, Pavel Obukhov · IOP Conference Series Materials Science and Engineering · 2021
Abstract The article is devoted to the development and analysis of methods of identifying dynamic objects. A system for identifying information from a luggage tag based on several neural networks with the SSD InceptionV2 architecture has been developed. These neural networks work with sufficiently high accuracy 82-95% and speed 7-10fps. Advantages and disadvantages of application of method of scale-invariant feature transform for identification of luggage tags are considered. The operability of the methods on real images has been tested.