Logo detection method based on target detection and metric learning

Abstract

The invention discloses a logo detection method based on target detection and metric learning. The logo detection method is characterized by specifically comprising the following steps: S1, constructing and training a logo detection model; s2, constructing and training a logo feature extraction model; and S3, detecting whether a candidate logo target exists in the to-be-detected picture or not, and determining whether the candidate logo target is a logo type in the logo retrieval picture library or not. According to the method, the logo position in the commodity poster data can be efficiently detected, the candidate region of the logo is determined, a new brand can be recognized, and redundancy detection is realized. The model does not need to be retrained, and the brand logo feature extraction accuracy can be more effectively improved. The system complexity is greatly simplified. Meanwhile, the recall rate is improved through logo detection, the accuracy is improved through logo feature extraction retrieval, and compared with a single target detection method, the recognition effect is better.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a logo detection method based on target detection and metric learning. Background technique [0002] With the development of online shopping, the infringement of product images on web pages has become more and more serious. For a large number of product pictures, manual review will consume a lot of manpower and material resources, so automatic logo infringement detection becomes very important. [0003] In the prior art, for the infringing logos, when the number of categories is small, the basic needs can be met through target detection, but when the categories of infringing logos continue to expand, it will become cumbersome to retrain the model each time , while wasting time, so the scalability of the logo detection model becomes important. Contents of the invention [0004] The technical problem to be solved by the present invention is to overcome the deficiencies of the pri...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, for the infringing logos, when the number of categories is small, the basic needs can be met through target detection, but when the categories of infringing logos continue to expand, it will become cumbersome to retrain the model each time , while wasting time, so the scalability of the logo detection model becomes very important

Images