Human body image generation model training method, human body image generation method and related devices
A technology for human image and model generation, which is applied in the computer field and can solve problems such as limited algorithm application scenarios, rough detail information, and unstable image generation models.
Active Publication Date: 2021-02-12
BEIJING XINTANG SICHUANG EDUCATIONAL TECH CO LTD
View PDF8 Cites 2 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] At present, the use of deep learning methods to generate images is only applied in the field of low-resolution images. When generating images with high resolution, the details of the generated images are very rough and the image generation model during training Unstable phenomenon, which severely limits the application scenarios of the algorithm
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0033] In the prior art, image generation by means of deep learning can only be applied in the field of low-resolution images, and the detailed information of images is relatively rough.
[0034] In one method, in order to improve the problem of insufficient details of the obtained image when generating high-resolution images, a method of local enhancement of distribution is used, that is: first, a CNN (Convolutional Neural Networks convolutional neural network) model is used to generate a complete The overall image; then, intercept the areas with insufficient details in the overall image, such as: hands, faces, etc., and then use the local CNN model to generate a local enhanced image, replace the corresponding area in the original image, and obtain the final overall image.
[0035] However, although the above method can improve the details of local areas to a certain extent, on the one hand, when there are multiple local areas with insufficient details, it is necessary to esta...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The embodiment of the invention provides a human body image generation model training method, a human body image generation method and related devices. The model training method comprises the steps ofobtaining a human body training image; fixing model parameters of each level of resolution generator lower than the current level of resolution, and using the current level of resolution generator togenerate a current level of overall generated image and a local generated image according to the current level of overall skeleton point image, acquiring current-level comprehensive loss by utilizingeach level of overall generated image, overall label image, locally generated image and local label image of which the resolution is lower than or equal to that of the current level, and optimizing model parameters of a current-level resolution generator to enable the current-level comprehensive loss to meet a loss threshold value; and adjusting the current-stage resolution, and optimizing the adjusted current-stage resolution generators until all the generators are optimized. According to the human body image generation model training method, the human body image generation method and the related devices provided by the embodiment of the invention, the human body image generation model precision and the human body image generation precision can be ensured.
Description
technical field [0001] Embodiments of the present invention relate to the computer field, and in particular, relate to a human body image generation model training method, device, device, and storage medium, and a human body image generation method, device, device, and storage medium. Background technique [0002] With the development of multimedia technology, in order to save cost and time, more and more images or videos of characters or cartoon characters are produced by means of image generation. [0003] At present, the use of deep learning methods to generate images is only applied in the field of low-resolution images. When generating images with high resolution, the details of the generated images are very rough and the image generation model during training Unstable phenomenon, which severely limits the application scenarios of the algorithm. [0004] Therefore, how to ensure the accuracy of the image generation model to achieve the accuracy of one-step image genera...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06T11/00G06T13/00
CPCG06T11/00G06T13/00G06T2207/20081G06T2207/20084G06T2207/30196
Inventor 冀志龙侯琦杨非
Owner BEIJING XINTANG SICHUANG EDUCATIONAL TECH CO LTD