Image processing method and device, storage medium and electronic equipment

Abstract

The invention relates to the technical field of deep learning, and provides a model search method and device, and an image processing method and device. Among them, the model search method includes: constructing the structure to be searched, at least one edge is connected between any two connected nodes of the structure to be searched, and each edge in the at least one edge corresponds to a different candidate operation; training the structure to be searched, in each In the first iteration, the model obtained after retaining one edge between each two connected nodes is determined as the model to be trained in this iteration. If the model contains edges that have been trained before, these edges are directly transferred. Trained parameters; after the structure to be searched is trained, at least one available model is selected from the models included in the structure to be searched according to the test results of the model performance. This method has high efficiency in searching models, and can cover a large search range to avoid missing valuable models.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a model search method and device, and an image processing method and device. Background technique [0002] Convolutional neural network is the most commonly used model in the field of deep learning. In recent years, researchers have continuously proposed convolutional neural network models with better performance through manual design and experiments, such as AlexNet, VGG16, Inception, Resnet, Xception etc. [0003] However, the manual design method has high requirements on the ability and experience of researchers, and it often takes a lot of time and energy for researchers to design an applicable model on a certain task or a certain data set. Moreover, there are many ways to design models, and the manual design method takes into account a very limited number of model structures, and it is likely to miss models with good performance. Contents of the invention ...

AI Technical Summary

Problems solved by technology

[0003] However, the manual design method has high requirements on the ability and experience of researchers, and it often takes a lot of time and energy for researchers to design an applicable model on a certain task or a certain data set.

Moreover, there are many ways to design models, and the manual design method takes into account the limited number of model structures, and it is likely to miss models with good performance.

Images