(SP)²Net for Generalized Zero-Label Semantic Segmentation


METADATA ONLY
Loading...

Date

2021

Publication Type

Conference Paper

ETH Bibliography

yes

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

Generalized zero-label semantic segmentation aims to make pixel-level predictions for both seen and unseen classes in an image. Prior works approach this task by leveraging semantic word embeddings to learn a semantic projection layer or generate features of unseen classes. However, those methods rely on standard segmentation networks that may not generalize well to unseen classes. To address this issue, we propose to leverage a class-agnostic segmentation prior provided by superpixels and introduce a superpixel pooling (SP-pooling) module as an intermediate layer of a segmentation network. Also, while prior works ignore the pixels of unseen classes that appear in training images, we propose to minimize the log probability of seen classes alleviating biased predictions in those ignore regions. We show that our (SP)2Net significantly outperforms the state-of-the-art on different data splits of PASCAL VOC 2012 and PASCAL-Context benchmarks.

Publication status

published

Book title

Pattern Recognition. DAGM GCPR 2021

Volume

13024

Pages / Article No.

235 - 249

Publisher

Springer

Event

43rd DAGM German Conference on Pattern Recognition (DAGM GCPR 2021)

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Scene understanding; Zero-shot semantic segmentation

Organisational unit

03514 - Van Gool, Luc (emeritus) / Van Gool, Luc (emeritus) check_circle

Notes

Conference lecture held on September 30, 2021.

Funding

Related publications and datasets