LEARNING METHOD AND LEARNING DEVICE FOR HETEROGENEOUS SENSOR FUSION BY USING MERGING NETWORK WHICH LEARNS NON-MAXIMUM SUPPRESSION

摘要

A learning method for generating integrated object detection information about an integrated image by integrating first object detection information and second object detection information is provided. The method includes (a) when the learning device acquires the first object detection information and the second object detection information, a concatenating network included in a deep neural network (DNN) causes the first original Generating a pair feature vector including information on a pair of the ROI and the second original ROI; (b) the learning device causes the discrimination network included in the DNN to apply a fully connected (FC) operation to the pair feature vector, thereby generating (i) discriminant vector and (ii) box regression vector. Creating; And (c) allowing the learning device to generate an integrated loss by the loss unit, and to learn at least some of the parameters included in the DNN by performing backpropagation using the integrated loss. Disclosed is a method comprising;