Visual perception and perception network training method, device, equipment and storage medium

Abstract

The present application provides a visual perception and perception network training method, device, device, and storage medium. The visual perception method uses the perception network to identify the acquired image to be perceived to determine the perception target and the pose state of the perception target, and finally The control instruction is determined according to the preset control algorithm and the pose state, so that the object to be controlled can determine a processing strategy for the perceived target according to the control instruction. It realizes the pose perception of the moving parts of the perception target, achieves the technical effect of refining the perception granularity, and deepening the analysis and understanding of the local motion of the object. The training method of the perceptual network obtains image data and model data, then uses the preset editing algorithm to generate edited images based on two-dimensional images and three-dimensional models, and finally trains the perceptual network to be trained according to the edited images and labels to determine the perceptual network . The technical effect of faster generation of more realistic training images with less domain difference is achieved.

Description

technical field [0001] The present application relates to the field of target detection, and in particular to a visual perception and perception network training method, device, device and storage medium. Background technique [0002] Object Detection is a basic problem in the field of computer vision. In uncontrolled natural scenes, quickly and accurately locating and identifying specific targets is an important functional basis for many artificial intelligence application scenarios. Visual perception is one application of object detection. [0003] The existing visual perception technology generally detects on the image through the visual perception system based on deep neural network, and obtains instance segmentation, that is, only the perceptual target can be obtained, and the overall pose estimation of the perceptual target is performed, which is called in the field The above situation is: relying on the bounding box and the contour to understand the object at a shall...

AI Technical Summary

Problems solved by technology

[0005] This application provides a visual perception and perception network training method, device, equipment and storage medium to solve the problem in the prior art that the perception granularity is too large, resulting in the inability to analyze and understand the local motion of objects

Images