Open access
Date
2023Type
- Conference Paper
ETH Bibliography
yes
Altmetrics
Abstract
Multi-task learning promises better model generaliza- tion on a target task by jointly optimizing it with an aux- iliary task. However, the current practice requires addi- tional labeling efforts for the auxiliary task, while not guar- anteeing better model performance. In this paper, we find that jointly training a dense prediction (target) task with a self-supervised (auxiliary) task can consistently improve the performance of the target task, while eliminating the need for labeling auxiliary tasks. We refer to this joint training as Composite Learning (CompL). Experiments of CompL on monocular depth estimation, semantic segmentation, and boundary detection show consistent performance improve- ments in fully and partially labeled datasets. Further analy- sis on depth estimation reveals that joint training with self- supervision outperforms most labeled auxiliary tasks. We also find that CompL can improve model robustness when the models are evaluated in new domains. These results demonstrate the benefits of self-supervision as an auxiliary task, and establish the design of novel task-specific self- supervised methods as a new axis of investigation for future multi-task learning research. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000592889Publication status
publishedExternal links
Book title
2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)Pages / Article No.
Publisher
IEEEEvent
Organisational unit
03514 - Van Gool, Luc / Van Gool, Luc
09688 - Yu, Fisher / Yu, Fisher
More
Show all metadata
ETH Bibliography
yes
Altmetrics