Real-time street view image semantic segmentation method based on deep multi-branch aggregation

Abstract

The invention discloses a real-time street view image semantic segmentation method based on deep multi-branch aggregation, and relates to a computer vision technology. A popular encoder-decoder structure is adopted; the method comprises the following steps: firstly, transforming a lightweight image classification network as a basis to serve as an encoder; dividing the encoder into different sub-networks, and sending features in each sub-network into a designed multi-branch feature aggregation network and a global context module; performing enhancement on spatial details and semantic information on features needing to be aggregated by using a lattice-type enhancement residual module and a feature transformation module in the multi-branch feature aggregation network; and finally, according to the sizes of the feature maps, aggregating the output feature maps of the global context module and the output feature maps of the multi-branch feature aggregation network step by step from small to large so as to obtain a final semantic segmentation result map. While the streetscape image with a large resolution is processed, high streetscape image semantic segmentation precision and real-time prediction speed are maintained.

Description

technical field [0001] The invention relates to computer vision technology, in particular to a method for semantic segmentation of real-time street scene images based on deep multi-branch aggregation. Background technique [0002] Semantic segmentation needs to assign pixel-level semantically interpretable categories to the target image, which plays a vital role in achieving complete scene understanding and is a very basic but challenging task in computer vision. In the past few years, due to the rise of autonomous driving systems and intelligent transportation systems, semantic segmentation of street view images has attracted more and more attention from experts and scholars in the field of computer vision. Generally speaking, these applications require fast interaction and response speed, so there is a strong demand for real-time street view image semantic segmentation algorithms. [0003] Traditional semantic segmentation methods mostly rely on artificially designed feat...

AI Technical Summary

Problems solved by technology

These complex deep neural networks usually require a lot of computational operations and memory consumption

Therefore, although these methods have achieved remarkable progress, their high computational cost and memory requirements also make it difficult to implement in real-world applications with limited computational resources (such as autonomous driving systems and driver assistance systems).

[0004] At present, many real-time semantic segmentation methods of street view images usually sacrifice a lot of spatial details or context information in order to obtain fast prediction speed, resulting in unacceptable segmentation accuracy.

Obviously, unlike the fast-developing semantic segmentation method of street view image that pursues high segmentation accuracy, the research on semantic segmentation method of street view image that meets real-time requirements without reducing too much accuracy is still lagging behind

Images