Weakly supervised image semantic segmentation method based on iterative mining of common features of objects

Abstract

The invention proposes a weakly supervised image semantic segmentation method based on iteratively mining common features of objects, which belongs to the technical field of pattern recognition. In the training phase, the method obtains a training data set, builds and trains a multi-label classification network, and obtains the initial seed area of ​​the object corresponding to each training image; then, obtains the superpixel area and area label of each training image for training area classification Network, get the updated region label of the superpixel region and use it to train the semantic segmentation network; after iteration, when the performance of the semantic segmentation network converges, the trained semantic segmentation network is obtained; in the use stage, the color image is input into the trained semantic segmentation network , the network outputs the semantic segmentation result of the image. The present invention can realize reliable pixel-level semantic segmentation under the condition of only image category labels, reduce the time and labor cost of data labeling, and has wide application prospects.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition, and in particular relates to a weakly supervised image semantic segmentation method based on iteratively mining common features of objects. Background technique [0002] Image semantic segmentation is an important research direction in pattern recognition and computer vision. It refers to the full understanding of image content through pixel-level recognition and segmentation of images. Therefore, it has a very broad application prospect in the fields of automatic driving and robot vision. . [0003] In intelligent application scenarios such as autonomous driving, the on-board computer needs to complete the perception of the scene first, and make corresponding movements and decisions based on the perceived information. Image semantic segmentation is an important part of visual scene perception. Image semantic segmentation realizes the computer's understanding of the pixel level of th...

AI Technical Summary

Problems solved by technology

[0006] However, under the condition that only image category labels are used as supervisory information, the difficulty of designing and training neural networks in existing weakly supervised image semantic segmentation is greatly increased.

Images