METHOD AND APPARATUS FOR GENERATING PSEUDO-LABEL DATA, METHOD AND APPARATUS FOR LEARNING DEEP NEURAL NETWORK USING PSEUDO-LABEL DATA

摘要

A pseudo-label data generating apparatus according to an embodiment of the present invention is pre-learned using data obtained through a first deep neural network and a second sensor that have been previously learned using data obtained through a first sensor. a memory in which a second deep neural network is stored; an input/output unit receiving first data and second data obtained from each of the first sensor and the second sensor that have been calibrated with each other; and using the data received from the input/output unit, selecting one of the first deep neural network and the second deep neural network, and inputting the first data or the second data into the selected deep neural network, A processor for generating pseudo-label data of an unselected deep neural network by using an output value output from the deep neural network may be included.