Lightweight network real-time semantic segmentation method based on attention mechanism

Abstract

The invention relates to a lightweight network real-time semantic segmentation method based on an attention mechanism, which is used for solving the problems that the segmentation precision and the segmentation efficiency are difficult to balance and the practical application cannot be met. The method comprises: preparing image data; and constructing a lightweight real-time semantic segmentation network based on an attention mechanism, providing a new asymmetric encoding and decoding network structure. In an encoder, a lightweight module-separable asymmetric module is used, and the module combines the advantages of depth separable asymmetric convolution and hole convolution. The calculation amount is greatly reduced while the precision is kept; an attention feature fusion module is designed in a decoder, features in the encoder and features in the decoder are fused, the fused features are selected and combined through an attention mechanism, the features useful for recovering image information are enhanced, and the network segmentation precision is effectively improved. Finally, semantic segmentation is achieved by using the trained segmentation network.

Description

Technical field: [0001] The invention belongs to the technical field of image semantic segmentation, and relates to a lightweight network real-time semantic segmentation method based on an attention mechanism. Background technique: [0002] In recent years, with the rapid development of computer technology and sensor technology, robot research has made great progress, and more and more service robots are widely used in social production and life. When a robot serves humans, it first needs to establish cognition and understanding of the surrounding environment, and then complete a series of other tasks, such as robot positioning, navigation, path planning, etc. Therefore, the cognition and understanding of the environment directly affects the performance of the robot. inferior. Semantic segmentation is a cornerstone technology of scene understanding. It groups each pixel in an image according to its semantic meaning, that is, it classifies each pixel in an image. After the ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a light-weight network real-time semantic segmentation method based on the attention mechanism, which can better solve the problem that the image semantic segmentation in the prior art is difficult to achieve a balance between segmentation accuracy and segmentation efficiency, so as to meet the needs of robots. The need for real-time segmentation in real-world environments

Images