Text and contour-based man-made controllable image synthesis method

Abstract

The invention discloses a text and contour-based man-made controllable image synthesis method, which comprises the following steps of: customized synthesis: drawing a basic contour and inputting basic text information, and encoding the basic contour and the basic text information to obtain respective feature vectors, combining the feature vectors of the basic contour and the basic text information together to synthesize a corresponding image; and optimization correction: inputting optimized text information, coding the synthesized image and the optimized text information to obtain corresponding feature vectors, and then combining the feature vectors of the synthesized image and the optimized text information together to synthesize and obtain an optimized image. According to the method, the controllability of image synthesis can be improved, and a high-quality image result can be synthesized under the condition of completely conforming to human subjective willingness.

Description

technical field [0001] The invention belongs to the technical field of image synthesis, in particular to an artificially controllable image synthesis method based on text and outlines. Background technique [0002] Controllable image synthesis is one of the most important technical problems of artificial intelligence in the field of computer vision. Compared with text and voice information, images contain richer content, which makes it difficult for machines to complete high-quality image synthesis. The rapid development of artificial intelligence in recent years has made a major breakthrough in image synthesis technology. At present, machines have been able to synthesize subjectively more realistic complex images (such as human faces, album covers, room layouts, etc.). With the advent of the era of artificial intelligence, machines have been given more missions, the most critical of which is to enable machines to understand human subjective thoughts more accurately so as t...

AI Technical Summary

Problems solved by technology

For example, if the category label "bird" is artificially input, the image synthesis technology based on the category label can synthesize an image of a bird, but the specific information of the bird (such as the color and size of the bird) is beyond human control

The method of inputting natural language description contains more information, so it can play a better role in human control, but this method lacks the constraints of the overall layout of the image, so that the synthetic results cannot better meet human expectations.

Images