Personalized Image Semantic Segmentation

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 PSS (Personalized 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 is available at https://github.com/zhangyuygss/PSS.

Introduction

Dataset License:

Our dataset is made available only for academic research. Although we have obtained the personalized photos’ copyright, the user’s privacy is still important. If you want to get access to our data, please send me a request from your school or company email. The request should include the purpose of using our dataset. Thank you for your understanding. （pt.jiang AT mail.nankai.edu.cn）

Citation

@inproceedings{zhang2021pss,
  title={Personalized Image Semantic Segmentation},
  author={Yu, Zhang and Chang-Bin, Zhang and Peng-Tao, Jiang and Ming-Ming, Cheng and Feng, Mao},
  booktitle={ICCV},
  year={2021}
}