Algorithm link optimization method and optimization system

Abstract

The invention discloses an optimization method and an optimization system of an algorithm link. The optimization method and the optimization system are used for improving the efficiency of an algorithm for iteratively repairing a training model. The method comprises the steps of generating a training mirror image file; generating a test mirror image file; obtaining a traceability instruction sent by operation and maintenance personnel, and tracing the training mirror image file and the test mirror image file according to the traceability instruction; deploying the first model to a self-training quick repair platform according to the training mirror image file, the test mirror image file and the deployment script; training the first model based on the production training data set to generate a second model; evaluating the first model to obtain a first algorithm precision score, and evaluating the second model to obtain a second algorithm precision score; determining a target model according to the first algorithm precision score and the second algorithm precision score; pushing the target model to an algorithm reasoning platform; and replacing the first model on the algorithm reasoning platform with the target model.

Description

technical field [0001] The present application relates to the field of computer science and technology, and in particular, to an optimization method and an optimization system for an algorithm link. Background technique [0002] In the production process of AI algorithms, the test model accuracy is often very high. However, in the customer's production environment, the accuracy of the algorithm's landing accuracy is not good. This is because the algorithm uses more general scene data as the training set in the development stage, and the algorithm accuracy can reach It is expected, however, in the actual environment, due to factors such as shooting angle, light intensity, and target moving, the actual landing effect of the algorithm after being installed in the customer environment is not good. Therefore, the algorithm model needs to be iteratively repaired. [0003] The iterative chain of enterprise-level production of computer vision algorithms generally includes 10 links: ...

AI Technical Summary

Problems solved by technology

[0002] In the production process of AI algorithms, the accuracy of the test model is often very high. However, in the customer's production environment, the accuracy of the algorithm's landing accuracy is not good. It is expected, but in the actual environment, due to factors such as shooting angle, light intensity, and moving targets, the actual landing effect of the algorithm after it is installed in the customer environment is not good, so it is necessary to iteratively repair the algorithm model

[0003] The enterprise-level production iteration link of computer vision algorithms generally includes 10 links: data labeling, code development, model training, model evaluation, model deployment, SDK packaging, streaming media packaging, algorithm testing, algorithm shelves, algorithm upgrades, generally speaking , Customized development and repair of an algorithm The full-link production cycle is 6 to 12 weeks, the cycle time is too long, the program is complicated, and it cannot meet the requirements of rapid iterative repair

Images