Image segmentation and classification method and device

Abstract

The invention relates to an image segmentation and classification method and device, and the method comprises the steps: S1, building an image segmentation and classification network model, and achieving the segmentation and classification of image frames, wherein the segmentation and classification network model comprises a context information extraction path and a spatial information extractionpath, the context information extraction path is in communication connection with the spatial information extraction path, and the context information extraction path outputs an image classification result; the spatial information extraction path outputs an image segmentation result; and S2, training the segmented and classified network model to enable the network model to reach a preset standard.According to the network model for segmentation and classification in the method, end-to-end real-time semantic segmentation can be performed on the video acquired by the mobile terminal, the targetstate category of the scene in the image frame can be identified in real time, and semantic segmentation of the video and classification of the scene state in the image frame are completed in parallel.

Description

technical field [0001] The invention relates to an image segmentation and classification method and device, and belongs to the technical field of image processing. Background technique [0002] For a video containing the target object to be analyzed, some image frames have the target object in the scene, and some image frames do not exist. At the same time, some image frames have the target object, but due to occlusion in the target area, or the target is not clear, or the target These image frames are unusable due to reasons such as the area is too small, or the pose of the target object is incorrect. Therefore, in order to realize the automatic photography of human-computer interaction, it is necessary to solve the problem of segmenting the target object area and giving the classification category of the frame image at the same time. [0003] The conventional method in the prior art is to divide the segmentation of the target object in the image frame and the classificati...

AI Technical Summary

Problems solved by technology

Moreover, most of the current segmentation deep learning models can only segment the target area, but cannot simultaneously classify and identify the validity of the scene state of the image frame, and cannot meet the needs of real-time analysis of the target object on the mobile electronic device side.

Although there is indeed a method in the field of autonomous driving that can simultaneously segment, classify, or detect image frames, the model used in this method takes up a large amount of memory, the calculation speed is slow, and it has high requirements for the hardware platform, which cannot meet the requirements of mobile devices. Requirements for video image processing in terminals

Images