IMAGE PROCESSING DEVICE AND ACTUATION METHOD AND ACTUATION PROGRAM THEREOF, OPERATION DEVICE AND ACTUATION METHOD AND ACTUATION PROGRAM THEREOF, AND MACHINE LEARNING SYSTEM

摘要

This image processing device is provided with: an extraction unit which, in an annotation image given as learning data to a machine learning model for implementing semantic segmentation involving differentiating multiple classes in the image by pixel unit, extracts, from multiple designated regions designated by class labels, a complex region which is at least part of the designated regions and which has a relatively complex outline; and a setting unit which sets an additional label for the complex region apart from the original labels originally designated in the annotation image.