Improving scene understanding through panoptic segmentation

A new approach makes object recognition more efficient by simultaneously performing foreground object segmentation and background scene segmentation in a single neural network.