Network training method and device, image processing method and device, storage medium and electronic equipment

Abstract

Embodiments of the invention provide a network training method and device, an image processing method and device, a storage medium and electronic equipment. The neural network training method comprises the following steps of: generating a neural network through parameters, obtaining an effect transformation parameter of a first sample image, and transforming the first sample image into a second sample image; obtaining effect classification and detection data of the second sample image through a classification neural network; and training the parameters to generate the neural network accordingto the effect classification and detection data of the second sample image and image effect classification labeling information of the first sample image. On the basis of a generative adversarial network, simple and object labeling data of image effect classification is taken as supervision information of training, accurate image effect parameter data labeling does not need to be carried out on selected sample images, and parameters for generating image effect transformation parameters are trained through a weak supervised learning manner so as to generate neural networks.

Description

technical field [0001] Embodiments of the present invention relate to image processing technologies, and in particular, to a training method of an image processing neural network, an image processing method, a device, a storage medium, and electronic equipment. Background technique [0002] In various image-related applications, the captured images need to be processed according to the needs of the application scene, for example, beautification processing, background blur processing, and cropping processing for portraits captured by mobile phones. [0003] Here, Image Aesthetic Enhancement (Image Aesthetic Enhancement) is an image enhancement technology that attempts to improve image quality or image aesthetics based on computer aesthetics. In the existing image aesthetic enhancement technology, a small number of sample images are mainly used for image feature extraction, and a model for generating enhanced parameters is trained based on a strongly supervised machine learnin...

AI Technical Summary

Problems solved by technology

[0004] However, in image effect processing tasks such as image aesthetic enhancement, it is extremely difficult to collect a considerable amount of high-quality training data and label image processing parameters, so the existing strongly supervised learning methods have great limitations.

Images