Multi-task Fusion for Efficient Panoptic-Part Segmentation

Sravan Jagadeesh, René Schuster, Didier Stricker

2023

Abstract

In this paper, we introduce a novel network that generates semantic, instance, and part segmentation using a shared encoder and effectively fuses them to achieve panoptic-part segmentation. Unifying these three segmentation problems allows for mutually improved and consistent representation learning. To fuse the predictions of all three heads efficiently, we introduce a parameter-free joint fusion module that dynamically balances the logits and fuses them to create panoptic-part segmentation. Our method is evaluated on the Cityscapes Panoptic Parts (CPP) and Pascal Panoptic Parts (PPP) datasets. For CPP, the PartPQ of our proposed model with joint fusion surpasses the previous state-of-the-art by 1.6 and 4.7 percentage points for all areas and segments with parts, respectively. On PPP, our joint fusion outperforms a model using the previous top-down merging strategy by 3.3 percentage points in PartPQ and 10.5 percentage points in PartPQ for partitionable classes.

Paper Citation

in Harvard Style

Jagadeesh S., Schuster R. and Stricker D. (2023). Multi-task Fusion for Efficient Panoptic-Part Segmentation. In Proceedings of the 12th International Conference on Pattern Recognition Applications and Methods - Volume 1: ICPRAM, ISBN 978-989-758-626-2, pages 15-26. DOI: 10.5220/0011616000003411

in Bibtex Style

@conference{icpram23,
author={Sravan Jagadeesh and René Schuster and Didier Stricker},
title={Multi-task Fusion for Efficient Panoptic-Part Segmentation},
booktitle={Proceedings of the 12th International Conference on Pattern Recognition Applications and Methods - Volume 1: ICPRAM,},
year={2023},
pages={15-26},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011616000003411},
isbn={978-989-758-626-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 12th International Conference on Pattern Recognition Applications and Methods - Volume 1: ICPRAM,
TI - Multi-task Fusion for Efficient Panoptic-Part Segmentation
SN - 978-989-758-626-2
AU - Jagadeesh S.
AU - Schuster R.
AU - Stricker D.
PY - 2023
SP - 15
EP - 26
DO - 10.5220/0011616000003411