Class learning for semantic segmentation from new knowledge

Bin Xie; Wenxiao Huang

doi:10.1117/12.3034914

11 July 2024 Class learning for semantic segmentation from new knowledge

Bin Xie, Wenxiao Huang

Proceedings Volume 13210, Third International Symposium on Computer Applications and Information Systems (ISCAIS 2024); 1321026 (2024) https://doi.org/10.1117/12.3034914
Event: Third International Symposium on Computer Applications and Information Systems (ISCAIS 2023), 2024, Wuhan, China

Abstract

Deep learning has achieved excellent results in semantic segmentation in computer vision, but this requires a large amount of data and computational power. Therefore, Class Incremental Semantic Segmentation (CISS) is becoming one of the current research hotspots, which involves gradually learning new classes in the incremental step to update the model. However, the existence of catastrophic forgetting can cause the model to forget the previously learned classes, especially when background class pixels are offset, which further exacerbates this problem in CISS. To address this issue, this article proposes a new approach that combines knowledge distillation and latent class comparison loss to alleviate catastrophic forgetting. The effectiveness of our method has been demonstrated through extensive experiments in existing CISS tasks and recently proposed challenging tasks.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Bin Xie and Wenxiao Huang "Class learning for semantic segmentation from new knowledge", Proc. SPIE 13210, Third International Symposium on Computer Applications and Information Systems (ISCAIS 2024), 1321026 (11 July 2024); https://doi.org/10.1117/12.3034914

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available