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

RISC-V-based TinyML target detection acceleration system and method, and storage medium

A RISC-V and target detection technology, applied in the field of deep learning, can solve problems such as difficult to further improve the computing speed, and achieve the effect of optimizing the arrangement structure

Pending Publication Date: 2022-05-24
SHANDONG INSPUR SCI RES INST CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] This application provides a TinyML target detection acceleration system, method, and storage medium based on RISC-V, which solves the technical problem that it is difficult to further improve the calculation speed when performing convolution calculations

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
  • RISC-V-based TinyML target detection acceleration system and method, and storage medium
  • RISC-V-based TinyML target detection acceleration system and method, and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0049] Embodiment 1: The ping-pong buffer unit contains two buffer areas, A and B, which can store two frames of images; the camera module uses a single camera, and the camera buffers the first frame image collected in the ping-pong buffer unit A area. The ping-pong buffer unit sends a signal that the buffering of area A is completed to the RISC-V soft core E906 unit; the camera buffers the second frame image collected in the ping-pong buffer unit B area. After the buffer is completed, the ping-pong buffer unit sends the RISC-V soft core E906 unit to the unit. Send the signal of completion of buffering in area B; the camera module enters the waiting state, when the camera receives the read completion signal in area A, it buffers the third frame image in area A, and sends the ping-pong buffer unit to the RISC-V soft core E906 unit Send the signal that the buffering in area A is completed; the camera module enters the waiting state. When the camera receives the reading completion...

Embodiment 2

[0051] Embodiment 2: The ping-pong buffer unit includes four buffer areas A, B, C, and D, and can buffer four frames of images; the camera module uses four cameras, and each camera occupies one buffer area.

[0052] After power-on, the four cameras collect the first frame of images, and store them in the four buffer areas of the corresponding ping-pong buffer unit. After the buffering is completed, send A, B, C, D buffer completion signal of four buffer areas.

[0053] Camera A enters the waiting state, and the RISC-V soft core E906 unit starts to read the data in the ping-pong buffer unit A area. After the reading is completed, it sends the camera A through the ping-pong buffer unit. A signal that the reading of the ping-pong buffer unit A area is completed; The reading of the ping-pong buffer unit A area is completed, the second frame data is buffered in the ping-pong buffer unit A area, and the RISC-V soft core E906 unit A area is notified that the buffering is completed; r...

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 RISC-V-based TinyML target detection acceleration system and method, and a storage medium, and the system comprises a ping-pong cache unit which is connected with a camera and an RISC-V soft core E906 unit, and is used for caching an image collected by the camera, and transmitting the image to the RISC-V soft core E906 unit; the RISC-V soft core E906 unit is connected with the ping-pong cache unit, the instruction cache unit and a data bus, and is used for reading and analyzing an instruction in the instruction cache unit, and controlling the weight and parameter cache unit, the convolution acceleration unit and the convolution cache unit to perform data processing through the data bus; and the convolution acceleration unit is connected with the data bus, the weight and parameter cache unit and the convolution cache unit, and is used for determining an image recognition result according to data in the weight and parameter cache unit and the convolution cache unit.

Description

technical field [0001] The present application relates to the field of deep learning, and in particular, to a RISC-V-based TinyML target detection acceleration system, method, and storage medium. Background technique [0002] Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-Power Microcontrollers (TinyML) is the intersection of machine learning and embedded IoT devices. and energy-saving computing and other fields have a very huge application potential. The current mainstream method is to use it in combination with ARM cores, but ARM cores are expensive to license. [0003] As an emerging instruction set architecture, the Reduced Instruction Set Computer (RISC-V) instruction set adopts the loose Berkeley Software Distribution (BSD) protocol, which has a strong open source atmosphere and can achieve complete autonomy. Controllable system design. [0004] Due to the complex convolution calculation and the large amount of calculation, although the speed of the ...

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): G06V10/94G06K9/62G06F12/0877G06V10/774
CPCG06F12/0877G06F18/214Y02D10/00
Inventor 王帅姜凯魏朝飞
Owner SHANDONG INSPUR SCI RES INST 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