Method for realizing IAE MEM post-processing picture identification and annotation based on deep migration

A technology of image recognition and depth, applied in the field of image processing, can solve the problems of difference in judgment results, low efficiency, lack of data annotation, etc., to achieve the effect of improving speed and reliability
CN110378436APending Publication Date: 2019-10-25四川升拓检测技术股份有限公司
3 Cites 7 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
四川升拓检测技术股份有限公司
Publication Date
2019-10-25

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a method for realizing IAE MEM post-processing picture identification and annotation based on deep migration. The method comprises the steps: A, collecting IAE MEM post-processing images, taking the collected images as a data set, and annotating the data set to define boundary frames of various defects in each image; B, constructing a Tiny Yolo model, generating a trainingset based on the data set, and training the Tiny Yolo model based on the training set; C, based on the trained Tiny Yolo model, identifying and annotating an IAE MEM post-processing picture to be identified. According to the method, the training effect which can only be achieved by big data can be achieved under the condition of small samples through deep transfer learning. The capacity of IAE MEMpost-processing intelligent recognition and labeling of various defects including ultra-thick, non-dense, under-thick and void defects is greatly improved.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The present invention relates to the field of image processing, in particular to a method for recognizing and labeling images after IAE MEM post-processing based on depth migration. Background technique

[0002] Modern society is in an era of big data. Every day and every hour, social networks, intelligent transportation, video surveillance, industry logistics, etc., all generate massive images, texts, voices and other data. The increase in data enables machine learning and deep learning models to rely on such massive amounts of data to continuously train and update the corresponding models, making the performance of the models better and better, and more suitable for specific scenarios. However, these big data have brought serious problems: there is always a lack of perfect data annotation.

[0003] As we all know, the training and updating of machine learning models depend on the labeling of data. However, although we can obtain massive amounts of ...

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