Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A Pixel Shader Based Deep Learning Model Execution Method

A technology of pixel shader and deep learning, which is applied in the field of deep learning model execution based on pixel shader, can solve the problems of complex hardware environment and no open source solutions, and reduce CPU load, facilitate enabling and disabling, and high The effect of compatibility

Active Publication Date: 2021-07-27
CHENDU PINGUO TECH CO LTD
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

For the Android platform, the general computing function of OpenGL ES 3.1 is used in the commercialization solution. The system requires Android 5.0 or higher, and the hardware environment is complex. Only less than 30% of the models can support it, and there is no open source available so far. Program

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
  • A Pixel Shader Based Deep Learning Model Execution Method
  • A Pixel Shader Based Deep Learning Model Execution Method
  • A Pixel Shader Based Deep Learning Model Execution Method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0035] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0036] In the present invention, the general calculation is abstractly expressed as: using a certain Operator (operator) to operate on the input Tensor (tensor) to obtain the result Tensor process.

[0037]For example: 1+5=6, 1 and 5 are two input Tensors, and the + operator is executed on these two Tensors, and the output Tensor: 6 will be obtained.

[0038] Another example: the Element-wise Add operation in the deep neural network is to add two Feature maps element by element, that is, input TensorA=[1, 2, 3, 4, 5], B=[4, 2, 3 , 5, 1], execute Element-wise AddOperator on A and B, and get the result Tensor: C=[5, 4, 6, 9, 6].

[0039] 1. Realize general computing based on pixel shader

[0040] The following is the entire process of realizing general-purpose com...

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 invention discloses a method for executing a deep learning model based on a pixel shader, including a. creating a Tensor buffer; b. loading and running an Operator; C. loading a deep learning model file; d. executing each hidden layer layer by layer, etc. Steps, where Operator, input Tensor, and output Tensor are used to represent operators, tensors, and results in general computing, respectively. The invention utilizes the GPU on the mobile phone that can only be used for pixel coloring to perform general calculations, thereby realizing the execution of the deep learning model accelerated by the GPU on the mobile phone.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a pixel shader-based deep learning model execution method. Background technique [0002] In recent years, relying on the advancement of artificial intelligence technology, image processing problems such as image content recognition and description, object recognition and tracking, face detection, and key point labeling can be better solved through deep learning, or in the original effect get promoted. However, the execution of the deep learning model requires a huge amount of computation, and it needs to rely on high-performance GPUs for acceleration. Therefore, it is very difficult to run the deep learning model on devices with limited computing capabilities such as mobile phones and tablets. Although most mobile phones currently on the market are equipped with GPUs, due to the limitations of the hardware itself or interface standards, they are usually only used for displaying i...

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 Patents(China)
IPC IPC(8): G06T15/00G06T1/40
CPCG06T1/20G06T15/005
Inventor 张靖淇徐滢
Owner CHENDU PINGUO TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com