Unlock instant, AI-driven research and patent intelligence for your innovation.

Training sample optimization method, target detection model generation method, equipment and medium

A technology for target detection and training samples, applied in biological neural network models, character and pattern recognition, instruments, etc., can solve problems such as low target detection accuracy, incomplete feature extraction of foreground target objects, and high computational complexity

Active Publication Date: 2021-04-23
NANJING TRANSWARP INTELLIGENCE CO LTD
View PDF8 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, the first form requires manual operation, and ignores the semantic information of the foreground target object, which may easily lead to incomplete feature extraction of the foreground target object or mixed interference information in the background, resulting in low target detection accuracy.
Although the second form improves the accuracy and robustness of target detection, target detection and semantic feature extraction are separated. It is necessary to add a semantic segmentation algorithm or model to extract semantic information separately on the basis of the target detection neural network. Only by combining the results of target detection and semantic segmentation can target detection under semantic rules be realized, which has high computational complexity and limited performance

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
  • Training sample optimization method, target detection model generation method, equipment and medium
  • Training sample optimization method, target detection model generation method, equipment and medium
  • Training sample optimization method, target detection model generation method, equipment and medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046] Figure 1a It is a flow chart of a method for optimizing training samples in a target detection model provided by Embodiment 1 of the present invention. This embodiment is applicable to the case of processing training samples when generating a target detection model. This method can be implemented in the target detection model The training sample optimization device is implemented, the device can be realized by software and / or hardware, and the device can be integrated in the computer, such as Figure 1a As shown, the method specifically includes:

[0047] Step 110, obtaining a training sample set, each training sample in the training sample set is labeled with a foreground target object and a background target object.

[0048] Wherein, the training samples include a large number of training samples. Figure 1b It is a schematic diagram of a training sample provided by Embodiment 1 of the present invention. Such as Figure 1b As shown, in the training samples (the first...

Embodiment 2

[0071] Figure 2a It is a flow chart of a method for generating a target detection model provided in Embodiment 2 of the present invention. This embodiment is applicable to the case of detecting a target when generating a target detection model. This method can be executed by a device for generating a target detection model , the device can be realized by software and / or hardware, and the device can be integrated in a computer, such as Figure 2a As shown, the method specifically includes:

[0072] Step 210, obtaining target optimized samples obtained after optimizing the training samples.

[0073] Wherein, the target optimization samples may be generated by the training sample optimization method in the target detection model provided in Embodiment 1 of the present invention. An object optimization sample may have a labeled box including a foreground object and a background object.

[0074] Step 220, using each target optimization sample, iteratively training the preset de...

Embodiment 3

[0107] image 3 It is a schematic structural diagram of a training sample optimization device in a target detection model provided by Embodiment 3 of the present invention. combine image 3 , the device includes: a training sample acquisition module 310 , a location constraint relationship determination module 320 and a label frame addition module 330 . in:

[0108] A training sample acquisition module 310, configured to acquire a training sample set, each training sample in the training sample set is marked with a foreground target object and a background target object;

[0109] The location constraint relationship determination module 320 is used to determine the semantic rule constraints between the foreground target object and the background target object according to the detection task of the target detection model;

[0110] The labeling frame adding module 330 is configured to obtain target optimization samples satisfying semantic rule constraints in each training sam...

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 embodiment of the invention discloses a training sample optimization method, a target detection model generation method, equipment and a medium. The training sample optimization method comprises the steps: acquiring a training sample set, wherein each training sample in the training sample set is marked with a foreground target object and a background target object; determining semantic rule constraints between the foreground target object and the background target object according to a detection task of the target detection model; and in each training sample, obtaining a target optimization sample meeting semantic rule constraints, and in each target optimization sample, generating a label box comprising a foreground target object and a background target object at the same time. According to the method, by adding the annotation boxes simultaneously including different types of targets, semantic information of the foreground target object is added in the sample training process, so that complete foreground target object features are extracted conveniently during target detection, the recognition accuracy is improved, and the calculation complexity is low.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of artificial intelligence, and in particular to a method for optimizing training samples, a method for generating a target detection model, equipment and media. Background technique [0002] In the field of artificial intelligence, for the detection of target objects, two forms of detection are usually used. The first form is to perform foreground target feature detection and recognition in the artificially designated detection area; the second form is to realize target detection through convolutional neural network in deep learning. [0003] However, the first form requires manual operation and ignores the semantic information of the foreground target object, which may easily lead to incomplete feature extraction of the foreground target object or mixed interference information in the background, resulting in low target detection accuracy. Although the second form improves the ac...

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): G06K9/00G06K9/32G06K9/34G06N3/04
Inventor 张燕夏正勋
Owner NANJING TRANSWARP INTELLIGENCE CO LTD