Method and device for generating training samples in nail semantic segmentation, equipment and medium

Abstract

The embodiment of the invention discloses a method and device for generating training samples in nail semantic segmentation, equipment and a medium. The method comprises the steps of acquiring a plurality of nail base plate images and a plurality of nail sample plate images, wherein the nail base plate images are nail images coated with primer of the same color or not coated, and the nail sample plate images are nail images coated with primer of multiple different colors; respectively marking the nail substrate image and the nail template image to determine a first nail area in the nail substrate image and a second nail area in the nail template image; and taking each nail substrate image as a template, and covering the first nail region in each nail substrate image by using the image of the second nail region in each nail template image to generate a new training sample. Therefore, a large number of training samples are directly generated through a small number of samples, the acquisition of the training samples is accelerated, the labor investment is reduced, and the acquisition difficulty of the training samples is also reduced.

Description

technical field [0001] The embodiments of the present invention relate to the field of deep learning technology, and in particular to a method, device, device and medium for generating training samples in nail semantic segmentation. Background technique [0002] With the development of deep learning, it has gradually been applied to the field of image recognition, and further deep learning can be used to realize the process of self-service nail art. However, to use deep learning for analysis, it is first necessary to train and test a good deep learning model through a large amount of sample data, so that the model can be used to obtain more accurate analysis results. [0003] The training data and test data of the current deep learning model are mostly collected one by one and manually calibrated, but the collection process is very troublesome and takes a long time, and the manual labeling process also takes a long time and Higher cost is a waste of manpower and material re...

AI Technical Summary

Problems solved by technology

[0003] The training data and test data of the current deep learning model are mostly collected one by one and manually calibrated, but the collection process is very troublesome and takes a long time, and the manual labeling process also takes a long time and effort. High cost, a great waste of manpower and material resources

At the same time, there are still situations where some samples are difficult to obtain, resulting in the inability to obtain sufficient training samples for model training

Images