Computer vision data set semi-automatic labeling method and system based on deep learning
A technology of computer vision and deep learning, applied in neural learning methods, computer components, computing, etc., can solve the problems of high difficulty in implementation, low efficiency of image labeling, and high labor costs
Inactive Publication Date: 2021-03-02
NARI TECH CO LTD
View PDF2 Cites 4 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0007] Purpose of the invention: In order to solve the problems of low image labeling efficiency in the electric power professional field, and the identification of defect types of electrical equipment requires the participation and guidance of business experts, resulting in high labor costs and high implementation difficulties. This invention proposes a computer vision data based on deep learning. Set semi-automatic labeling method and system, by using this method, semi-automatic labeling can be quickly realized, and the human-machine process of artificial intelligence assisting humans can be realized
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0042] Now further set forth the technical scheme of the present invention.
[0043] Such as figure 1 A semi-automatic labeling method for computer vision datasets based on deep learning is shown, including the following steps:
[0044] Step 1: Collect the image set A of the target sample to be identified, and randomly sample it at a ratio of 10% to generate the image set B;
[0045] Step 2: Perform manual labeling, image preprocessing and image expansion on image set B to improve the quality of the sample training set; specifically include:
[0046] In the image preprocessing stage, including but not limited to the following measures: (1) image geometric correction completed according to the internal parameters of the image acquisition device; (2) in order to highlight the information of interest in the image, weaken or remove unnecessary information to make useful The information is strengthened, which is easy to distinguish or explain, and image color enhancement processi...
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
Login to View More Abstract
The invention provides a computer vision data set semi-automatic labeling method and system based on deep learning, and the method comprises the following steps: 1, carrying out the random sampling ofa sample picture set A according to a certain proportion, and generating a picture set B; 2, performing manual labeling, image expansion and image preprocessing on the picture set B subjected to theoperation in the step 1 to generate a training set B; 3, performing initial training on a deep neural network model in the Faster-rcnn algorithm by adopting the training set B to obtain a model weight; 4, marking the sample picture set A by using the model weight and combining a man-machine coupling marking method, and carrying out image expansion and image preprocessing on the marked sample picture set A to generate a training set A; 5, retraining the deep neural network model in the Faster-rcnn algorithm by adopting the training set A; and 6, inputting a to-be-identified image into the deepneural network model trained in the step 5 to obtain a labeling result.
Description
technical field [0001] The invention belongs to the technical field of computer vision data labeling, in particular to a method and system for semi-automatic labeling of computer vision data sets based on deep learning. Background technique [0002] Artificial intelligence is the most popular technology at present, but how to efficiently preprocess existing data is a key part of artificial intelligence technology. In the branch of computer vision, it is necessary to label images and obtain image datasets required for deep neural network training. However, the work of image annotation has the following problems: [0003] (1) The high repetition rate leads to heavy workload and cumbersome work details, consuming a lot of time and energy of software algorithm engineers, and virtually increasing labor costs; [0004] (2) In professional business fields such as electric power or medical treatment, the work of image annotation must rely on the prior knowledge of business personn...
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
Login to View More IPC IPC(8): G06N3/04G06N3/08G06K9/62G06T5/00
CPCG06N3/08G06T2207/20048G06T2207/20182G06N3/045G06F18/24G06F18/214G06T5/73G06T5/70
Inventor陈天宇徐弘升张琪培李子琪陆继翔杨志宏
OwnerNARI TECH CO LTD