Neural network training and image processing methods and devices, computer readable storage medium and electronic equipment

A neural network training, neural network technology, applied in image data processing, image enhancement, image analysis and other directions, can solve problems such as limitation, low accuracy of foreground and background segmentation, and low processing speed.

Inactive Publication Date: 2018-10-16
BEIJING SENSETIME TECH DEV CO LTD
View PDF4 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the existing foreground and background segmentation processing technology has the problem of low foreground and background segmentation accuracy or low processing speed. When the processing speed is limited, the foreground and background segmentation accuracy will also be limited, resulting in the final monocular Poor bokeh effect

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Neural network training and image processing methods and devices, computer readable storage medium and electronic equipment
  • Neural network training and image processing methods and devices, computer readable storage medium and electronic equipment
  • Neural network training and image processing methods and devices, computer readable storage medium and electronic equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0049] figure 1 is a flow chart showing a neural network training method according to Embodiment 1 of the present invention.

[0050] refer to figure 1 , in step S110, the first convolution processing is performed on the training sample image by the neural network system to obtain the first segmentation result of the foreground and background segmentation of the training sample image; and, the second convolution processing is performed on the first segmentation result to obtain the training sample image. The second segmentation result of the foreground-background segmentation of the sample image.

[0051] In the embodiment of the present invention, the training sample images are used to train a neural network system that performs foreground and background segmentation processing on images. The training sample images can be images captured by any camera in any scene. Optionally, the neural network system is a deep neural network system. The setting of the specific structure...

Embodiment 2

[0062] figure 2 is a flow chart showing a neural network training method according to Embodiment 2 of the present invention.

[0063] refer to figure 2 , in step S210, the first convolution processing is performed on the training sample image through the first convolution sub-network of the neural network system, and the first segmentation result of the foreground and background segmentation of the training sample image is obtained.

[0064]In the embodiment of the present invention, the neural network system is used to segment the foreground and background of the image, and the neural network system includes a first convolution sub-network, which is used to perform the first convolution processing on the image to obtain the first segmentation result. Here, the first convolution sub-network may include a plurality of convolution layers, and the first convolution process includes convolution operations performed by the plurality of convolution layers respectively. In the fi...

Embodiment 3

[0085] Figure 4 is a flowchart illustrating an image processing method according to Embodiment 3 of the present invention.

[0086] refer to Figure 4 , in step S410, the image to be processed is input into the neural network system, and the foreground and background segmentation processing is performed on the image to be processed through the neural network system.

[0087] In the embodiment of the present invention, the neural network system is used to perform foreground and background segmentation processing on the image, specifically, it may be a neural network system trained by the neural network training method in Embodiment 1 or Embodiment 2 of the present invention.

[0088] Optionally, the first convolution processing is performed on the image to be processed by the neural network system to obtain the first segmentation result of the foreground and background segmentation of the image to be processed; and the second convolution processing is performed on the first s...

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

No PUM Login to view more

Abstract

The embodiment of the invention provides neural network training and image processing methods and devices, a computer readable storage medium and electronic equipment. The neural network training method comprises the steps that first convolution processing is carried out on a training sample image through a neural network system, a first segmentation result of foreground background segmentation ofthe training sample image is obtained, second convolution processing is carried out on the first segmentation result, and a second segmentation result of foreground background segmentation of the training sample image is obtained; according to the difference between a labeling segmentation result of the foreground background segmentation of the training sample image and the first segmentation result and the difference between the labeling segmentation result and the second segmentation result, the neural network system is trained. By means of the technical scheme, the image foreground background segmentation accuracy can be effectively improved, and the high processing speed is guaranteed so that the effect of monocular vision blurring processing can be effectively improved.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of computer vision, and in particular to a neural network training method, device, computer-readable storage medium, and electronic equipment, and an image processing method, device, computer-readable storage medium, and electronic equipment. Background technique [0002] The background blur of the image can make the subject clearly displayed, which is very popular among photography enthusiasts. Since the blurring effect of the image is mainly realized by using the principle of optical imaging and using a large lens aperture on the hardware, the blurring function of the image is mainly integrated on professional cameras such as SLR cameras. The mobile terminal device is limited and can only be installed with a small aperture lens. When the user takes a photo with the mobile terminal device, he can only generate an image with no blur effect or only a weak blur effect. [0003] Curre...

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
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/194
CPCG06T7/194G06T2207/10004G06T2207/20081G06T2207/20084
Inventor 王嘉
Owner BEIJING SENSETIME TECH DEV CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap