Evaluation and calling method and system for face recognition algorithm engine

Abstract

The invention provides an evaluation and calling method and system for a face recognition algorithm engine. The method comprises the following steps: converting a calling mode of an algorithm engine and an output content format of the algorithm engine into a unified calling mode and a unified content format; carrying out face recognition by using different algorithm engines through snapshot images obtained by monitoring equipment; performing accuracy scoring on the algorithm engine according to the matching hit condition; and performing performance scoring on the algorithm engine according to the time consumption and the comparison failure rate of the algorithm engine, performing quantitative calculation on the accuracy and the performance of the algorithm engine, performing weighted summation by using the accuracy score and the performance score to obtain a comprehensive score of the algorithm engine, and generating an evaluation report of the algorithm engine; and evaluating and calling the engines according to the comprehensive score. The advantages of each engine are fully played to achieve reasonable calling of the algorithm engine.

Description

technical field [0001] The invention relates to the field of algorithm engine evaluation, in particular to an evaluation and calling method and system for a face recognition algorithm engine. Background technique [0002] At present, most face detection systems mainly use a face recognition algorithm engine for face recognition. Generally, the face is recognized after taking a photo of a static picture, and then compared with the preset face map. The system is largely limited by the accuracy of the selected face recognition algorithm engine. If the accuracy of the face recognition algorithm engine is not high, it will affect the use of the face detection system. [0003] The current algorithm engine has the following problems: the scheduling method and the format of the defined data content are not uniform, there is no long-term and online dynamic evaluation method for the accuracy of the comparison results, and it is impossible to achieve integrated acquisition of various a...

AI Technical Summary

Problems solved by technology

The system is largely limited by the accuracy of the selected face recognition algorithm engine. If the accuracy of the face recognition algorithm engine is not high, it will affect the use of the face detection system.

[0003] The current algorithm engine has the following problems: the scheduling method and the format of the defined data content are not uniform, there is no long-term and online dynamic evaluation method for the accuracy of the comparison results, and it is impossible to achieve integrated acquisition of various algorithm engines to achieve normalized output. It is impossible to evaluate the engine capability online based on the normalized results

Images