Depth detection method and device, equipment and storage medium

Abstract

The invention provides a depth detection method, device and equipment and a storage medium, relates to the field of artificial intelligence, in particular to the field of computer vision and deep learning, and can be applied to intelligent robots and automatic driving scenes. According to the specific implementation scheme, high-level semantic features in a to-be-detected image are extracted, wherein the high-level semantic features are used for representing a target object in the to-be-detected image; the high-level semantic features are input into a pre-trained depth estimation branch network to obtain the distribution probability of the target object in each sub-interval of a depth prediction interval; and the depth value of the target object is determined according to the distribution probability of the target object in each sub-interval and the depth value represented by each sub-interval. According to the technology disclosed by the invention, the prediction task of the depth value can be converted into the classification task through the designed depth estimation branch network of adaptive depth distribution, the finally obtained depth value is relatively accurate, and the 3D positioning precision can be improved in the application of 3D object detection for the image.

Description

technical field [0001] The present disclosure relates to the field of artificial intelligence, specifically to the fields of computer vision and deep learning, and can be applied to intelligent robots and automatic driving scenarios. Background technique [0002] Monocular 3D detection mainly depends on the key point prediction of the 3D object projected onto the 2D image, and then by predicting the 3D attributes (length, width, height) and the depth value of the object, the real 3D bounding box of the object is restored to complete the 3D detection task. [0003] In related technologies, for depth prediction, the head branch network is usually used to predict the depth value of the object alone, which has the defect of low accuracy, thus affecting the performance of 3D detection. Contents of the invention [0004] The disclosure provides a depth detection method, device, equipment and storage medium. [0005] According to an aspect of the present disclosure, a deep dete...

AI Technical Summary

Problems solved by technology

[0003] In related technologies, for depth prediction, the head branch network is usually used to predict the depth value of the object alone, which has the defect of low accuracy, thus affecting the performance of 3D detection

Images