A Guided Semantic Segmentation Method Based on Image Boundary Knowledge Transfer

Abstract

The invention discloses a guided semantic segmentation method based on the transfer of image boundary knowledge, which includes the following steps: S1, the construction of the guided segmentation module, and S2, the construction of constraint conditions, which are specifically divided into three parts, namely supervision of limited samples and representation consistency constraints , and boundary consistency constraints; S3, boundary knowledge transfer module construction, which mainly includes pseudo triple discriminant data generation, boundary confrontation learning; the guided semantic segmentation method based on image boundary knowledge transfer established through the above steps can Ten labeled samples and a large amount of task-independent image data, using image boundary knowledge, specifying semantically related images, and using semantic space feature constraints, to achieve semantic segmentation of specific types of sample targets.

Description

technical field [0001] The invention relates to the field of small-sample semantic segmentation, in particular to a guided semantic segmentation method based on transfer of image boundary knowledge. Background technique [0002] Deep neural networks have achieved remarkable results in many computer vision applications, such as image semantic segmentation tasks that play an important role in autonomous driving and medical imaging. Generally speaking, training a deep neural network requires a large amount of labeled sample data, but the process of data acquisition and labeling is often time-consuming and labor-intensive. In response to this problem, existing work mainly focuses on the two tasks of small sample learning and transfer learning. Few-shot learning aims to learn to train a reliable model through few labeled samples, while transfer learning is to migrate a model learned on a task to a new scene. Although some progress has been made in small sample learning and tran...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problem that a large number of complex labeled samples are required in the prior art, and at the same time it is difficult to specify specific semantics for segmentation, and to achieve the purpose of semantic segmentation of specific types of sample targets, the present invention adopts the following technical solutions:

Images