A multi-region license plate fast training method using a large model to reuse a basic sample

Abstract

The application discloses a kind of multi-region license plate fast training methods for utilizing large model multiplexing basic sample, it is related to license plate fast training technical field, including: after the original vehicle picture set is not qualified picture cleaning and standardization preprocessing, multidimensional information is marked, and basic sample set is obtained;Pre-construct multiple parameterized license plate templates of multiple target area license plates;With the multiple parameterized license plate templates as input, license plate intelligent fusion of the basic sample set is carried out, and multiple training sample sets are generated;Using the multiple training sample sets, license plate recognition model training of the multiple target area license plates is carried out.The application solves the technical problems that the existing technology requires repeated collection and sample labeling for different area license plates, leading to long sample preparation period and low model training efficiency, achieving the technical effect of improving the sample preparation efficiency and training efficiency of multi-region license plate recognition model.

Images