Image semantic segmentation network training method, device and equipment and storage medium

Abstract

The invention discloses an image semantic segmentation network training method, device and equipment and a storage medium, and relates to the field of image semantic segmentation. The method comprisesthe steps of training a first image semantic segmentation network according to a training image set, wherein the training images in the training image set contain the annotation information; performing mask processing on the network weight of the first image semantic segmentation network through a mask network to obtain a second image semantic segmentation network, the mask network being used forscreening network weights insensitive to image feature distribution in the first image semantic segmentation network; and according to the training image set and the test image set, training the second image semantic segmentation network, the distribution of the training image set being different from that of the test image set. In the embodiment of the invention, as the mask network can filter the network weight sensitive to the feature distribution, the second image semantic segmentation network screened by the network weight also has better generalization on the test image set, thereby improving the test accuracy of the test image set.

Description

technical field [0001] The embodiments of the present application relate to the field of image semantic segmentation, and in particular to a training method, device, equipment and storage medium for an image semantic segmentation network. Background technique [0002] Semantic image segmentation is a technique that distinguishes different objects contained in an image and identifies the category to which each object belongs. In the field of artificial intelligence, image semantic segmentation networks are usually trained based on convolutional neural networks. [0003] In the related technology, after the initial image semantic segmentation network is constructed based on the convolutional neural network, the initial image semantic segmentation network is firstly trained using the training image set, and after the network training is completed, the trained image semantic segmentation is performed using the test image set The network is tested to determine the image semantic...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a training method, device, equipment and storage medium for an image semantic segmentation network, which can solve the problem of using the test image set to test the image semantic segmentation network when the distribution of the training image set and the test image set are different. The problem of low accuracy of the test results obtained

Images