Method and device for training 6D attitude estimation network based on depth learning iterative matching

Abstract

The present disclosure relates to a 6D attitude estimation network training method and apparatus based on depth learning iterative matching. The method comprises: based on the three-dimensional modelof the target object and the initial 6D attitude estimation, obtaining a rendered picture of the target object and a first partition mask, the picture will be rendered, first partition mask, the observation picture of the target object and the second segmentation mask of the target object in the observation picture are inputted into the depth convolution neural network, obtaining a 6D attitude estimate, a third segmentation mask and an optical stream, updating the initial 6D attitude estimate with the obtained 6D attitude estimate, replacing the second segmentation mask with a third segmentation mask, and re-performing the steps to iteratively train the depth convolutional neural network. The training method proposed in the embodiments of the present disclosure does not need to rely on depth information when improving the initial 6D attitude estimation, and the estimation result is accurate. It is robust to illumination, occlusion and other problems, and can deal with textured and non-textured objects simultaneously.

Description

technical field [0001] The present disclosure relates to the field of artificial intelligence, in particular to a 6D pose estimation network training method and device based on deep learning iterative matching. Background technique [0002] Obtaining the pose of an object in 3D space from a 2D image is very important in many real-world applications. For example, in the field of robotics, recognizing the 6D pose of an object, that is, the 3D position and 3D orientation of an object, can be used for tasks such as grasping or motion planning. Provides key information; in virtual reality scenes, accurate 6D object poses allow people to interact with objects. [0003] In traditional technologies, depth cameras are generally used for object pose estimation. However, depth cameras have many limitations, such as limitations in frame rate, field of view, resolution, and depth range, making it difficult for these technologies that rely on depth cameras to detect small, transparent, o...

AI Technical Summary

Problems solved by technology

[0005] In view of this, this disclosure proposes a 6D pose estimation network training method and device based on deep learning iterative matching to solve the problem that the 6D pose estimation of objects obtained by existing deep learning methods is not accurate enough, and lacks a depth-independent Information can improve the method of 6D attitude estimation

Images