Adaptive data enhancement method for target detection
A target-oriented, adaptive technology, applied in the fields of instruments, character and pattern recognition, computer parts, etc., can solve the problems of difficult detection of small objects, unbalanced classification, high sensitivity of training batches, etc., to reduce batch sensitivity The effect of improving the richness of content, alleviating classification imbalance, and improving content richness
Pending Publication Date: 2022-01-14
SHANGHAI NORMAL UNIVERSITY
View PDF0 Cites 2 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] The purpose of the present invention is to provide an adaptive data enhancement method for target detection in order to overcome the defects of small object detection difficulty, high sensitivity of training batches and unbalanced classification in the prior art
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 1
[0041] Such as figure 1 As shown, this embodiment provides an adaptive data enhancement method for target detection, including the following steps:
[0042] Traverse each picture in the data set, execute adaptive area removal algorithm, adaptive object selection algorithm and adaptive annotation filtering algorithm, and obtain the enhanced data set;
[0043] The adaptive area removal algorithm includes: obtaining the picture to be enhanced from the data set, traversing the area in the picture to be enhanced, and removing the area whose aspect ratio is unbalanced and whose area is smaller than the preset area threshold;
[0044] The adaptive object selection algorithm includes: according to the classification quantity information of a single object and the relationship between objects, select the object area from the data set to fill in the area to be removed in the image to be enhanced;
[0045] The adaptive annotation filtering algorithm includes: according to the bounding b...
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 relates to an adaptive data enhancement method for target detection, which comprises the following steps of: executing the following algorithms on a to-be-enhanced picture: an adaptive region removal algorithm: acquiring the to-be-enhanced picture from a data set, traversing the regions in the to-be-enhanced picture, and removing the region of which the length-width ratio is unbalanced and the area is smaller than a preset area threshold; an adaptive object selection algorithm: according to the classification number information of a single object and the relationship between the objects, selecting an object region from the data set and filling the removed region in the to-be-enhanced picture with the object region; and an adaptive label filtering algorithm: filtering out an object which is not suitable for being used as a model learning target according to the bounding box information of the original object in the to-be-enhanced picture. Compared with the prior art, the method has the advantages that the content richness of the data set can be effectively improved, and the network trained based on the enhanced target detection data set has model improvement on the problems of small target detection, batch sensitivity reduction, classification imbalance relief and the like.
Description
technical field [0001] The invention relates to the technical field of data enhancement, in particular to an adaptive data enhancement method oriented to target detection. Background technique [0002] Object detection (object detection) is an image processing technology, which has a very wide range of applications in actual reality. However, the current target detection algorithm still has the following problems in the actual use process: first, it is difficult to detect small objects, that is, the trained model has low recognition accuracy for small objects; second, it is sensitive to training batches Larger, that is, the effect of the model training process changes significantly with the change of the batch size, which will cause the instability of model learning; finally, the data set used for target detection model training often has the problem of classification imbalance, that is The number of objects in different categories varies significantly. Contents of the in...
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 Patent Type & Authority Applications(China)
IPC IPC(8): G06V10/24G06K9/62
CPCG06F18/214
Inventor 林晓孙树州黄伟郑晓妹黄继风蒋林华
Owner SHANGHAI NORMAL UNIVERSITY