A method and device for centralized storage of eigenvalues ​​for multiple target recognition algorithms

Abstract

The invention discloses a method and device for centralized storage of feature values ​​for various target recognition algorithms. The method includes: acquiring a picture to be recognized, and using a target recognition algorithm to identify the picture to be recognized, so as to generate a Real-time structured data and real-time feature values ​​corresponding to the picture to be recognized; according to the algorithm name and algorithm version of the target recognition algorithm, the implemented structured data and real-time feature values ​​are stored in the first storage space and the second storage space respectively ; Based on the same target algorithm, similarity matching is performed between the real-time feature value and the standard picture feature value pre-stored in the third storage space, and a matching result is generated; according to the matching result, a corresponding application interface is called. The invention can optimize, adjust and centrally store the characteristic values ​​of multiple target recognition algorithms, and develop a unified application interface at the front end, realize unified management and maintenance of multiple target recognition algorithms, and improve the flexibility of the application interface.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for centralized storage of feature values ​​for various target recognition algorithms. Background technique [0002] With the rapid development of big data, artificial intelligence, deep learning and other technologies and methods, target recognition and detection technology has also entered the fast lane of development, and has become the hottest research hotspot in the field of artificial intelligence today. The current target recognition technology is mainly based on the convolutional neural network algorithm to generate vector values ​​that can be understood by computers, that is, eigenvalues, and further detection, comparison, and recognition methods. With the rapid change of technology, there are more and more target recognition algorithms based on different application scenarios. Different manufacturers and different algorithms produce differen...

AI Technical Summary

Problems solved by technology

On the other hand, since an algorithm corresponds to a specific application interface, when adding an algorithm or modifying a certain feature value comparison SDK, the corresponding interface must be modified accordingly. The application interface not only lacks flexibility, but also increases the development workload.

Images