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

Implementation method of convolutional neural network based on heterogeneous FPGA (Field Programmable Gate Array) and fused with multiple resolutions

A convolutional neural network and multi-resolution technology, which is applied in the field of convolutional neural network based on heterogeneous FPGA platform and fusion multi-resolution, to achieve the effect of improving system performance, high precision, and overcoming low precision

Active Publication Date: 2022-04-15
JIANGSU UNIV
View PDF4 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, due to the huge amount of calculation of the convolutional neural network, it is necessary to make a trade-off between speed and accuracy when using the FPGA platform to implement the convolutional neural network model.

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
  • Implementation method of convolutional neural network based on heterogeneous FPGA (Field Programmable Gate Array) and fused with multiple resolutions
  • Implementation method of convolutional neural network based on heterogeneous FPGA (Field Programmable Gate Array) and fused with multiple resolutions
  • Implementation method of convolutional neural network based on heterogeneous FPGA (Field Programmable Gate Array) and fused with multiple resolutions

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach

[0068] refer to figure 1 , is a flow chart of an implementation method based on a heterogeneous FPGA and a fusion multi-resolution convolutional neural network according to an embodiment of the present invention. The specific implementation is as follows:

[0069] Step 1: Convolutional Neural Network (CNN) algorithm model fusion with multi-resolution, the embodiment of the present invention is described by taking YOLO-V2 algorithm fusion with multi-resolution model as an example.

[0070] refer to figure 2 , for the multi-resolution YOLO-V2 improved model - Multi-resolution YOLO-V2 model structure diagram.

[0071] By designing its passthrough structure as a series connection of high and low resolution networks, the recognition ability of the overall network is enhanced. The advantage of this difference from the single use of high-resolution networks is that the high-resolution network is designed as a front-end network with a passthrough structure. The number of layers of...

Embodiment approach

[0109] refer to Figure 4 , is a flow chart of the processing method of the embodiment of the present invention, and the specific implementation is as follows:

[0110] (1) Get image

[0111] The PS side (processing system, ARM) collects rice images through the camera.

[0112] (2) Image preprocessing on the PS side

[0113] First, normalize the image, divide the input RGB image by 256, so that each pixel value is in the interval [0, 1].

[0114] Then the resulting image is converted to a size of 416×416, and the insufficient padding constant is filled with a value of 0.5.

[0115] Store the resulting image in DDR.

[0116] (3) Run high and low resolution networks in parallel

[0117] As the main control unit, the PS side first starts the operation of the high-resolution network after image preprocessing is completed, and then starts the operation of the scaling module and the low-resolution network.

[0118] The low-resolution network is actually formed by the PS side con...

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 an implementation method of a convolutional neural network based on a heterogeneous FPGA and fusing multiple resolutions, and the method comprises the following steps: 1, fusing multiple resolutions through a convolutional neural network algorithm model, and fusing a multi-resolution model through a YOLO-V2 algorithm; 2, a YOLO-V2 model with high resolution is used for training; 3, weight parameters are recombined and quantized; and a fourth step of realizing a Multi-solution YOLO-V2 algorithm on a heterogeneous FPGA (Field Programmable Gate Array) platform through hardware and software. According to the invention, a multi-resolution fusion technology is provided, the YOLO-V2 algorithm is improved by using the technology, and the detection capability of the network is greatly improved under the condition that the speed is hardly lost.

Description

technical field [0001] The invention relates to the field of target detection, and specifically refers to an implementation method based on a heterogeneous FPGA platform and a fusion multi-resolution convolutional neural network. Background technique [0002] Convolutional neural network is currently the most widely used network model in the field of target detection. It is a deep learning technology evolved from multi-layer perceptron (MLP). Due to its structural characteristics of local area connection and weight sharing , and its learning and work are carried out end-to-end, which makes the convolutional neural network perform well in the field of image processing; at the same time, the structure of the convolutional neural network is flexible, and designers can construct the most suitable network structure according to their own needs. It is very beneficial to the realization of detection tasks. [0003] Currently commonly used hardware platforms are CPU, GPU and FPGA. ...

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): G06V10/774G06V10/77G06V10/80G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/213G06F18/25G06F18/214Y02D10/00
Inventor 徐雷钧姚沛东白雪
Owner JIANGSU UNIV
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