Multi-task Fusion for Efficient Panoptic-Part Segmentation

TL;DR

This paper presents a new multi-task network that unifies semantic, instance, and part segmentation for panoptic-part segmentation, using a parameter-free fusion module to improve accuracy and efficiency.

Contribution

The paper introduces a novel shared encoder network with a parameter-free joint fusion module for panoptic-part segmentation, achieving state-of-the-art results.

Findings

Outperforms previous methods on Cityscapes Panoptic Parts dataset.

Achieves higher PartPQ scores on Pascal Panoptic Parts dataset.

Efficient fusion improves segmentation consistency and accuracy.

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.