The paper introduces Panoptic FPN, a single-network architecture designed to perform both instance segmentation and semantic segmentation tasks simultaneously. The approach leverages the Mask R-CNN framework with a Feature Pyramid Network (FPN) backbone, adding a semantic segmentation branch to the existing instance segmentation branch. This design allows the network to generate region-based outputs for instance segmentation and dense-pixel outputs for semantic segmentation. The authors demonstrate that Panoptic FPN achieves high accuracy in both tasks, outperforming separate networks for each task and serving as a robust baseline for future research in panoptic segmentation. The method is memory and computationally efficient, making it suitable for large-scale datasets like COCO and Cityscapes. The paper also includes detailed experimental results and ablation studies to validate the effectiveness of the proposed approach.The paper introduces Panoptic FPN, a single-network architecture designed to perform both instance segmentation and semantic segmentation tasks simultaneously. The approach leverages the Mask R-CNN framework with a Feature Pyramid Network (FPN) backbone, adding a semantic segmentation branch to the existing instance segmentation branch. This design allows the network to generate region-based outputs for instance segmentation and dense-pixel outputs for semantic segmentation. The authors demonstrate that Panoptic FPN achieves high accuracy in both tasks, outperforming separate networks for each task and serving as a robust baseline for future research in panoptic segmentation. The method is memory and computationally efficient, making it suitable for large-scale datasets like COCO and Cityscapes. The paper also includes detailed experimental results and ablation studies to validate the effectiveness of the proposed approach.