Depth estimation method combined with semantic edge

Abstract

The invention relates to a semantic edge-combined depth estimation method. The method comprises the following steps of: acquiring an image of which the depth is to be estimated; inputting the image into a trained deep learning network to obtain a depth prediction map and a semantic edge prediction map; the deep learning network comprises a shared feature extraction module, a depth estimation module, an edge enhancement weight module, a depth edge semantic classification module and a semantic edge detection module; the shared feature extraction module is used for extracting feature information in the image and transmitting the feature information to the depth estimation module and the semantic edge detection module; the depth estimation module guides parallax smoothing through the semantic edge output by the semantic edge detection module, and performs depth estimation in an image double-reconstruction mode; the edge enhancement weight module forms a feature result needing to be fused by the semantic edge detection module based on the depth edge of the depth prediction map output by the depth estimation module; the depth edge semantic classification module is used for performing depth edge semantic classification prediction; and the semantic edge detection module is used for outputting semantic edge classification prediction of the image. The accuracy can be improved.

Description

technical field [0001] The present invention relates to the technical field of computer vision, in particular to a depth estimation method combined with semantic edges. Background technique [0002] Depth estimation and semantic edge extraction are fundamental problems in computer vision, and the results of their tasks can be deployed in practical sciences such as autonomous driving, virtual reality, robotics, etc., to assist in achieving better results. Depth estimation refers to parsing 3D perceptual information from an image. The semantic edge extraction task is a task that combines image edge extraction and classification tasks, which can simultaneously obtain the semantic information of the edge and the boundary of the object. Currently, the method of deep learning is used. to handle both types of tasks. [0003] Depth estimation is further divided into monocular depth estimation and multi-camera depth estimation. Monocular depth estimation has the advantages of fast p...

AI Technical Summary

Problems solved by technology

The problem with the above methods is that there is no truth value of additional information during inference, so we can only use the trained method to generate the required additional information, thus losing the reliability of using the truth value to guide during inference

The problem with the above method of using dynamic weights is that the weights are forcibly learned from features, lacking theoretical support, and adding a learning layer with more parameters, which makes the network complex

However, the above two methods heavily rely on the accuracy of a given label and are both step-by-step patterns, which may be suboptimal and inefficient

Images