System and Method for Preprocessing and Data Set Augmentation for training with Data Processing

摘要

The present invention relates to an apparatus and a method for data set expansion generation and preprocessing for artificial intellectual (AI) training used in a three-dimensional (3D) data processing, capable of enhancing the utilization of data and the precision of an output result by generating a larger number of data sets for training AI from a smaller number of data. The apparatus includes: a data input unit to receive a first number of 3D data; a sub-data selecting unit to randomly select sub-data from one piece of data including several pieces of sub-data within a predetermined range; a sub-data sequence generating unit to perform expansion and generation of 3D data in second number larger than the first number by generating a sequence of the sub-data; a motion history image (MHI) two-dimensional (2D) data generating unit to transform a plurality of 3D data, which is generated in the sub-data sequence generating unit, into two-dimensional (2D) data through an MHI procedure; and a machine learning processing unit to extract a feature map by extracting a feature of the 2D data by receiving, as an input, the 2D of the MHI 2D data generating unit, applying a filter to the input, outputting a feature map having a data amount reduced by extracting representative values, and connecting all made features with each other for learning the same.