High-resolution street view picture semantic segmentation training and real-time segmentation method

Abstract

The invention discloses a training method and a using method of an image semantic segmentation model. The training method comprises the steps that training images marked with semantic segmentation information in advance are input into a feature extraction module of a network; the feature extraction module combines the two advantages of the high processing speed of a low-resolution picture and the high deduction quality of a high-resolution picture, and a feature map obtained through calculation is output; then the feature map is sent into a segmentation module for deconvolution, and the feature map is restored to 1 / 4 size of the original map; the type weight of each pixel is marked to obtain a prediction result; and finally, the parameters of the network are corrected according to the prediction information of the trained image and the information marked in advance. The use method is similar to the training method, and the last 1 / 4 size of image is upsampled and recovered to the original image size. According to the segmentation method, the calculated amount and the consumed time are greatly reduced, and the segmentation method can run at the speed of 30 frames under the high resolution of 1024 * 2048, and meanwhile the high-quality inference effect is achieved.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a semantic segmentation training and real-time segmentation method of high-resolution street view pictures. Background technique [0002] Semantic segmentation of images is a very important field in computer vision. It refers to identifying images at the pixel level, that is, marking the object category to which each pixel in the image belongs, which can deepen the machine's understanding of the scenes, objects, and characters in the picture. understand. This technology has broad application prospects in the field of autonomous driving and medical field. [0003] With the application of convolutional neural networks in recent years, the field of semantic segmentation has made great progress. The most mainstream solutions for image semantic segmentation are mainly based on convolutional neural network (CNN), which learns various semantic feature tables contained in ima...

AI Technical Summary

Problems solved by technology

[0006] Purpose of the invention: Aiming at the problem that the real-time performance improvement of semantic segmentation is not obvious, a fast semantic segmentation network with high accuracy is proposed to improve the real-time performance of semantic segmentation of images

Images