Learning from Extrinsic and Intrinsic Supervisions for Domain Generalization
Other conference paper

Altmetrics Information

Other information
AbstractThe generalization capability of neural networks across domains is crucial for real-world applications. We argue that a generalized object recognition system should well understand the relationships among different images and also the images themselves at the same time. To this end, we present a new domain generalization framework (called EISNet) that learns how to generalize across domains simultaneously from extrinsic relationship supervision and intrinsic self-supervision for images from multi-source domains. To be specific, we formulate our framework with feature embedding using a multi-task learning paradigm. Besides conducting the common supervised recognition task, we seamlessly integrate a momentum metric learning task and a self-supervised auxiliary task to collectively integrate the extrinsic and intrinsic supervisions. Also, we develop an effective momentum metric learning scheme with the K-hard negative mining to boost the network generalization ability. We demonstrate the effectiveness of our approach on two standard object recognition benchmarks VLCS and PACS, and show that our EISNet achieves state-of-the-art performance.
All Author(s) ListWang S., Yu L., Li C., Fu CW., Heng PA.
Name of ConferenceECCV 2020
Start Date of Conference23/08/2020
End Date of Conference28/08/2020
Place of ConferenceUK
Country/Region of ConferenceGreat Britain
Proceedings TitleLecture Notes in Computer Science
Volume Number12354
Pages159 - 176
LanguagesEnglish-United Kingdom
KeywordsDomain generalization, Unsupervised learning, Metric learning, Self-supervision

Last updated on 2021-21-04 at 23:30