FPGA-based three-dimensional convolver

A three-dimensional convolution and convolutional technology, applied in the field of video processing, can solve the problems of low parallelism, increased data transmission bandwidth, multiple loading of input data, etc., so as to improve data input efficiency and increase operation speed. , the effect of improving computational efficiency

Inactive Publication Date: 2017-11-28
XIDIAN UNIV
View PDF2 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This structure occupies less resources, but the degree of parallelism is not high, and the input data needs to be loaded multiple times, which increases the requirements for the data transmission bandwidth between the data source and the FPGA, thereby limiting the overall speed of the three-dimensional convolution operation.

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
  • FPGA-based three-dimensional convolver
  • FPGA-based three-dimensional convolver
  • FPGA-based three-dimensional convolver

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0028] The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.

[0029] Such as figure 1 As shown, the FPGA-based three-dimensional convolution device includes an input data storage module 1, a two-dimensional convolution device array module 3, an adder module 5 and a data output controller module 6 implemented in the FPGA, wherein:

[0030] The input data storage module 1 is used to cache multiple feature maps to be processed in the internal RAM of the FPGA and send them, taking three feature maps to be processed as an example.

[0031] The two-dimensional convolution device array module 3 includes multiple two-dimensional convolution devices, which are used to convolve multiple feature maps to be processed with two-dimensional convolution kernels, and output multiple convolution results. The embodiment requires two-dimensional The number of convolvers is 9.

[0032] The adder module 5 includes ...

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 FPGA-based three-dimensional convolver, and aims at solving the technical problem that input data is repeatedly loaded in the prior art. The FPGA-based three-dimensional convolver comprises six modules realized in an FPGA: an input data storage module, a three-dimensional convolution kernel storage module, a two-dimensional convolver array module, an intermediate data delay line module, a summer module and a data output controller module, wherein the input data storage module is used for carrying out input cache on a to-be-processed feature map; the three-dimensional convolution kernel storage module is used for storing a three-dimensional convolution kernel in the FPGA in a form of a plurality of two-dimensional convolution kernels; the two-dimensional convolver array module is used for convolving the to-be-processed feature map and the three-dimensional convolution kernel and then outputting the convolved to-be-processed feature map and three-dimensional convolution kernel; the intermediate data delay line module is used for adding convolution results of appointed two-dimensional convolver, delaying the adding result and then outputting the adding result; the summer module is used for adding output of the two-dimensional convolver array module and output of an intermediate data delay line and then carrying out output; and the data output controller module is used for controlling the sequence of a plurality of outputs of the summer module. The FPGA-based three-dimensional convolver is capable of remarkably enhancing the three-dimensional convolution operation speed and can be used for the aspect of target tracking or behavior detection.

Description

technical field [0001] The invention belongs to the technical field of video processing, and relates to an FPGA-based three-dimensional convolution device, which can be used in a system based on an FPGA hardware platform for target tracking or human behavior detection using convolutional neural network technology. Background technique [0002] With the advancement of microelectronics manufacturing technology and integrated circuit design, Field Programmable Gate Array (FPGA) has been developed rapidly, because of its rich resources, flexible design, and low power consumption, it has been widely used. The FPGA implementation platform has broad development prospects because of its high performance, low power consumption, and reconfigurable characteristics. As a programmable standard device, FPGA can be applied to various functions, which requires a short development cycle, low development cost, and strong flexibility. In recent years, the development of high-level synthesis t...

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): G06G7/19G06F17/15
CPCG06G7/19G06F17/153
Inventor 王海赵伟沈越邵梦君王珂王中正
Owner XIDIAN UNIV
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