4 HD LEARNING METHOD AND LEARNING DEVICE FOR ATTENTION-DRIVEN IMAGE SEGMENTATION BY USING AT LEAST ONE ADAPTIVE LOSS WEIGHT MAP TO BE USED FOR UPDATING HD MAPS REQUIRED TO SATISFY LEVEL 4 OF AUTONOMOUS VEHICLES AND TESTING METHOD AND TESTING DEVICE USING THE SAME

摘要

Attention-driven image segmentation method using at least one adaptive loss weight map can be used to update the HD map required to meet level 4 of the autonomous vehicle. In this way, blurry objects such as lanes and road markings seen from a distance can be more accurately detected. In addition, in the military where PIA identification is important, the method can be usefully performed to distinguish aircraft markings or military uniforms from a distance. In the above method, the learning apparatus causes the softmax layer to generate a softmax score; Causing the loss weight layer to generate a prediction error value, and applying a loss weight operation thereto to generate a loss weight value; And causing a softmax loss layer to generate an initial softmax loss value generated by referring to the softmax score and a corresponding GT, and an adjusted softmax loss value by referring to the loss weight value. Is provided.