Image collection expansion method, device, equipment and readable storage medium

Abstract

The invention discloses an image set expansion method. By determining the current enhancement strategy to be called from the preset enhancement strategy library; calling the Bayesian optimization algorithm to determine the parameters to be tested, and calling the Bayesian optimization algorithm can be from countless enhancement strategies. Quickly determine several parameters with excellent performance enhancement performance in the call parameters, and determine the optimal parameters among the parameters to be tested. The optimal parameter is the parameter with the best enhancement effect among the parameters to be tested; The image enhancement strategy and optimal parameters of the original image set can ensure the high availability of the generated image, so the currently obtained image set can be directly used as the output data set without occupying too many resources for image validity verification, which can quickly Solve the problem of insufficient number of image sets, and can reduce the implementation cost. The invention also discloses an image collection expansion device, equipment and a readable storage medium, which have the above beneficial effects.

Description

technical field [0001] The invention relates to the field of data management, in particular to an image set expansion method, device, equipment and readable storage medium. Background technique [0002] In the field of image classification (such as the current model classification, face skin color classification, etc.), the image data set is the most critical part. Insufficient labeled image data often leads to overfitting and sampling noise, which means Then the model will fail to generalize to undiscovered examples. This problem can be alleviated by image augmentation. Image augmentation is to artificially generate new images by performing image transformations (such as rotation, cropping, occlusion, scaling, etc.) on the original image set. Image augmentation can effectively increase the number and diversity of images seen by the network. [0003] For image augmentation, it is very important how to call the appropriate enhancement algorithm to expand the existing image ...

AI Technical Summary

Problems solved by technology

At present, the traditional method uses reinforcement learning to optimize the image set. Due to the wide search range, this method will occupy a lot of resources, and the requirements for computing power are too high. The implementation cost is extremely high, and it is not suitable for low-cost usage scenarios.

Images