Method and device for generating training data

Abstract

The invention discloses a training data generation method and device. The method comprises the steps of obtaining original point cloud data of a road surface; Determining a point corresponding to theroad traffic sign as a target point from the original point cloud data; Carrying out region division on the original point cloud of the road surface by taking a target point corresponding to the roadsurface traffic sign as a reference to obtain a sample region set which comprises a positive sample region and a negative sample region, wherein at least one target point exists in the positive sampleregion, and no target point exists in the negative sample region; And according to the point cloud data falling into the sample area in the original point cloud data, generating a grey-scale map corresponding to the sample area, the grey-scale map of the positive sample area serving as a positive example sample for training the pavement traffic sign recognition model, and the grey-scale map of the negative sample area serving as a negative example sample for training the pavement traffic sign recognition model. Manual manufacturing is not needed, manpower resources are saved, and it can be guaranteed that enough training samples can be generated by controlling the number of divided sample areas.

Description

technical field [0001] The present application relates to the technical field of training data generation, and more specifically, to a method and device for generating training data. Background technique [0002] When using machine learning to train various road traffic sign recognition models, a large amount of training sample data is required to ensure the recognition accuracy of the trained road traffic sign recognition model. [0003] The existing training sample data is generally produced manually, that is, the pictures of the road surface are manually recognized, and the pictures containing road traffic signs are used as positive samples, and the pictures that do not contain road traffic signs are used as negative samples. [0004] Obviously, the method of manually producing training data requires a lot of human resources, and the number of training samples generated is not large enough to guarantee the recognition accuracy of the trained road traffic sign recognition ...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the present application provides a method and device for generating training data to solve the problems that the existing manual training data consumes human resources and the number of training samples produced is insufficient

Images