Device and method for quickly generating hand key point data set

Abstract

The invention discloses a device and method for quickly generating a hand key point data set, and belongs to the field of computer vision. The device comprises an arithmetic unit, a linear guide rail,a simulation hand model and an image acquisition unit, and the simulation hand model and the image acquisition unit are arranged on the linear guide rail, wherein the simulation hand model is used for sensing three-dimensional coordinates of each key point of the hand relative to the reference point; the image acquisition unit is used for acquiring the length and height of the reference point relative to the image acquisition unit; the operation unit is used for calculating spatial three-dimensional coordinates of each key point relative to the image acquisition unit according to the three-dimensional coordinates, length and height of each key point relative to the reference point; and calculating the two-dimensional image coordinates of each key point under the image coordinate system according to the mapping relation between the spatial three-dimensional coordinates and the image coordinates. According to the method, the hand key point data set is quickly and accurately generated, so that automatic labeling is realized, the shielded key points can be accurately labeled, and the defects of low manual labeling speed and subjective assumption of key point positions are overcome.

Description

technical field [0001] The invention belongs to the field of computer vision, and more specifically relates to a device and method for rapidly generating hand key point data sets. Background technique [0002] With the development of technology, the application of augmented reality (Augmented Reality, AR) and virtual reality (Virtual Reality, VR) and other technologies is gradually popularized, and people can interact with 3D images in virtual scenes through finger movements. The realization of these functions is based on the sensitive capture of finger movements to accurately obtain the gesture information of each joint of the finger. [0003] In order to capture the posture information of each joint of the finger, a deep learning network is usually used to train the algorithm model. The deep learning network requires a large number of hand datasets with key points marked as network training samples. At present, the manual annotation method is usually used to annotate the...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the defects and improvement needs of the prior art, the present invention provides a device and method for quickly generating hand key point datasets, the purpose of which is to automatically generate hand key point datasets quickly and accurately, so as to realize automatic labeling, and It can accurately mark the occluded key points, which solves the problems of low efficiency, time-consuming and inaccurate standards of manual marking

Images