Label-free six-dimensional object attitude prediction method and device based on reinforcement learning

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a label-free six-dimensional object attitude prediction method and device based on reinforcement learning. According to the technical scheme, the method comprises: obtaining a to-be-predicted target image, wherein the target image is a two-dimensional image comprising a target object; according to the targetimage, carrying out attitude prediction by adopting a pre-trained attitude prediction model to obtain a prediction result, wherein the attitude prediction model is a model obtained by carrying out reinforcement learning according to a sample image; and determining a three-dimensional position and a three-dimensional direction of the target object according to the prediction result. According to the embodiment of the invention, the attitude prediction model is trained by introducing reinforcement learning, and according to the target image, the attitude prediction model obtained by pre-trainingis adopted to perform attitude prediction, so that the problem of six-dimensional object attitude estimation based on the two-dimensional image can be solved under the condition of no real attitude annotation, and the prediction effect of label-free six-dimensional object attitude prediction is ensured.

Description

Technical field [0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a method and device for unlabeled six-dimensional object posture prediction based on reinforcement learning. Background technique [0002] The six-dimensional object pose prediction method includes a method of locating the three-dimensional position and three-dimensional direction of a target object from a two-dimensional image. [0003] In the related art, the six-dimensional object pose prediction method may include the following steps: a computer device obtains a two-dimensional image to be detected, compares the rendered object and the observed object of the target object in the two-dimensional space, and predicts the difference between the current estimated value and the true value. The six-dimensional attitude difference is used to estimate the six-dimensional object attitude. [0004] The above-mentioned methods involve real posture labeling, and real...

AI Technical Summary

Problems solved by technology

However, the acquisition process of 2D images annotated with real 6D object poses is quite complicated, which greatly limits the implementation of the above methods.

At present, a suitable and accurate method for predicting the pose of six-dimensional objects has not been provided

Images