Sample augmentation method for target key point detection

Abstract

The invention discloses a sample augmentation method for target key point detection. Firstly, an original sample for target key point detection is obtained, wherein the original sample is composed ofan image and annotation information of a target key point corresponding to the image; then, based on the image and target key point labeling information of the original sample and parameters of augmentation operation, gray values of pixel points of the image of the augmentation sample and corresponding unit value coordinates of labeled target key points in the image of the original sample in the image of the augmentation sample are determined. By means of the method, semantic alignment of the image of the augmented sample and the image of the original sample can be achieved, and high-precisionsample augmentation is achieved. The method and device are especially suitable for target key point detection algorithms which are extremely sensitive to augmentation precision.

Description

technical field [0001] The present application relates to the technical field of computer image processing, and in particular to a sample augmentation technique for object key point detection. Background technique [0002] In the field of computer image processing technology, a series of similar but different training samples are obtained through sample augmentation operations, that is, by randomly changing the acquired samples, thereby expanding the diversity of the training data set. Especially for the neural network model based on deep learning, the augmentation operation can reduce the dependence of the neural network model on certain attributes, thereby improving the generalization ability of the neural network model. [0003] During the operation of the existing sample augmentation methods, due to the irrationality of the algorithm design, the augmented sample and the original sample are not semantically aligned. The error caused by this sample augmentation will affec...

AI Technical Summary

Problems solved by technology

[0003] During the operation of the existing sample augmentation method, due to the irrationality of the algorithm design, the augmented sample and the original sample are not aligned at the semantic level

The error caused by this sample augmentation will affect the convergence speed of the neural network model training process, and will also affect the performance of the neural network model during the test process, especially the target key point detection algorithm that is extremely sensitive to augmented accuracy. neural network model

Images