A learning method and learning device for improving segmentation performance used in detecting events including pedestrian events, automobile events, falling events, and fallen events using edge loss, and a test method and test device using the learning method and learning device.

摘要

To provide a learning method for improving segmentation performance to be used for detecting events such as a pedestrian event, an automobile event, a falling event, and a fallen event.SOLUTION: A learning method for improving segmentation performance includes steps of: allowing k convolutional layers to generate k encoded feature maps; allowing k-1 deconvolutional layers to sequentially generate k-1 decoded feature maps; allowing h mask layers to refer to h basic decoded feature maps outputted from h deconvolutional layers corresponding thereto and h edge feature maps generated by extracting edge parts from the h basic decoded feature maps; and allowing h edge loss layers to generate h edge losses by referring to the edge parts and GTs corresponding to this.SELECTED DRAWING: Figure 6