Personalized Image Semantic Segmentation

TL;DR

This paper introduces the problem of personalized image segmentation, presents a new dataset called PIS, surveys recent related work, and proposes a baseline method that leverages inter-image context to improve segmentation accuracy on personalized images.

Contribution

It is the first to address personalized image segmentation, providing a new dataset, a survey of existing methods, and a baseline approach utilizing inter-image context.

Findings

Our method outperforms existing approaches on the PIS dataset.

The PIS dataset enables future research in personalized segmentation.

Survey results highlight the gap in current personalized segmentation techniques.

Abstract

Semantic segmentation models trained on public datasets have achieved great success in recent years. However, these models didn't consider the personalization issue of segmentation though it is important in practice. In this paper, we address the problem of personalized image segmentation. The objective is to generate more accurate segmentation results on unlabeled personalized images by investigating the data's personalized traits. To open up future research in this area, we collect a large dataset containing various users' personalized images called PIS (Personalized Image Semantic Segmentation). We also survey some recent researches related to this problem and report their performance on our dataset. Furthermore, by observing the correlation among a user's personalized images, we propose a baseline method that incorporates the inter-image context when segmenting certain images. Extensive experiments show that our method outperforms the existing methods on the proposed dataset. The code and the PIS dataset will be made publicly available.