Real-time target detection method for computing resource limited platform deployment

A technology for target detection and computing resources, applied in the fields of deep learning and image processing, to achieve the effect of increasing depth and width, ensuring accuracy, and requiring less communication

Active Publication Date: 2019-08-09
匀熵科技无锡有限公司
View PDF4 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

How to achieve real-time and accurate performance of YOLO on embedded and mobile devices, and realize real-time monitoring of multiple objects is still a major challenge

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
  • Real-time target detection method for computing resource limited platform deployment
  • Real-time target detection method for computing resource limited platform deployment
  • Real-time target detection method for computing resource limited platform deployment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0043] The technical solution of the present invention is: a method for object detection based on image processing, said method comprising the following steps:

[0044] (1) Deploy Tinier-YOLO on Jetson TX1, and collect images through the camera;

[0045] (2) Tinier-YOLO reads the images collected by the camera, where,

[0046] The Tinier-YOLO is an improved YOLO-v3-tiny network structure: the alternate operation of the first five convolutional layers and pooling layers of the YOLO-v3-tiny network structure is reserved, and then the Fire modules in the five SqueezeNet are connected sequentially , output to the first pass-through layer, and then the pass-through layer is connected to the Fire module in the sixth SqueezeNet, and uses the Dense connection to connect the output feature maps of the five Fire modules to the input of the sixth Fire module, and the sixth The data of the Fire module is output to the second pass-through layer and a 1*1 bottleneck layer, and the subseque...

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 real-time target detection method for computing resource limited platform deployment, and belongs to the field of deep learning and image processing. According to the methodthe YOLO-v3-tiny neural network is improved. The Tinier-YOLO reserves the first five convolution layers and pooling layers of the YOLO-v3-tiny and the prediction of two different scales, a Fire modulein the SqueezeNet is introduced, and a 1 * 1 bottleneck layer is connected with the Dense, so that the structure can be operated on an embedded AI platform in real time. The model size of the Tinier-YOLO provided by the invention is only 7.9 MB, and the real-time performance is improved by 21.8% compared with that of YOLO-v3-tiny, and is improved by 70.8% compared with that of YOLO-v2-tiny; compared with YOLO-v3-tiny, the accuracy of the method is improved by 10.1%, and compared with YOLO-v2-tiny, the accuracy of the method is improved by about 18.2%. According to the Tinier-YOLO provided bythe invention, the purpose of real-time detection still can be realized on a platform with limited computing resources, and the effect is better.

Description

technical field [0001] The invention relates to a real-time target detection method for platform deployment with limited computing resources, belonging to the fields of deep learning and image processing. Background technique [0002] Object detection is an important task in many emerging fields, such as robot navigation, autonomous driving, etc. In these complex scenes, object detection methods based on deep learning methods have greater advantages than traditional methods, and object detection algorithms based on deep learning continue to rise, such as R-CNN, SPPNet, fast-R-CNN, faster-R- CNN, R-FCN, and FPN. Although these object detection algorithms have achieved unprecedented accuracy, the detection speed is not fast, which is far from meeting the real-time requirements on devices with low computing power. At the same time, the size of deep learning models usually takes up a lot of storage space and requires powerful GPU computing capabilities. However, in most practi...

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): G06K9/00G06N3/04
CPCG06V20/10G06N3/045
Inventor 方伟任培铭王林孙俊吴小俊
Owner 匀熵科技无锡有限公司
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