Saved in:
Bibliographic Details
Main Authors: Houyon, Joachim, Cioppa, Anthony, Ghunaim, Yasir, Alfarra, Motasem, Halin, Anaïs, Henry, Maxim, Ghanem, Bernard, Van Droogenbroeck, Marc
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2304.01239
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916617944825856
author Houyon, Joachim
Cioppa, Anthony
Ghunaim, Yasir
Alfarra, Motasem
Halin, Anaïs
Henry, Maxim
Ghanem, Bernard
Van Droogenbroeck, Marc
author_facet Houyon, Joachim
Cioppa, Anthony
Ghunaim, Yasir
Alfarra, Motasem
Halin, Anaïs
Henry, Maxim
Ghanem, Bernard
Van Droogenbroeck, Marc
contents In recent years, online distillation has emerged as a powerful technique for adapting real-time deep neural networks on the fly using a slow, but accurate teacher model. However, a major challenge in online distillation is catastrophic forgetting when the domain shifts, which occurs when the student model is updated with data from the new domain and forgets previously learned knowledge. In this paper, we propose a solution to this issue by leveraging the power of continual learning methods to reduce the impact of domain shifts. Specifically, we integrate several state-of-the-art continual learning methods in the context of online distillation and demonstrate their effectiveness in reducing catastrophic forgetting. Furthermore, we provide a detailed analysis of our proposed solution in the case of cyclic domain shifts. Our experimental results demonstrate the efficacy of our approach in improving the robustness and accuracy of online distillation, with potential applications in domains such as video surveillance or autonomous driving. Overall, our work represents an important step forward in the field of online distillation and continual learning, with the potential to significantly impact real-world applications.
format Preprint
id arxiv_https___arxiv_org_abs_2304_01239
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Online Distillation with Continual Learning for Cyclic Domain Shifts
Houyon, Joachim
Cioppa, Anthony
Ghunaim, Yasir
Alfarra, Motasem
Halin, Anaïs
Henry, Maxim
Ghanem, Bernard
Van Droogenbroeck, Marc
Computer Vision and Pattern Recognition
Machine Learning
In recent years, online distillation has emerged as a powerful technique for adapting real-time deep neural networks on the fly using a slow, but accurate teacher model. However, a major challenge in online distillation is catastrophic forgetting when the domain shifts, which occurs when the student model is updated with data from the new domain and forgets previously learned knowledge. In this paper, we propose a solution to this issue by leveraging the power of continual learning methods to reduce the impact of domain shifts. Specifically, we integrate several state-of-the-art continual learning methods in the context of online distillation and demonstrate their effectiveness in reducing catastrophic forgetting. Furthermore, we provide a detailed analysis of our proposed solution in the case of cyclic domain shifts. Our experimental results demonstrate the efficacy of our approach in improving the robustness and accuracy of online distillation, with potential applications in domains such as video surveillance or autonomous driving. Overall, our work represents an important step forward in the field of online distillation and continual learning, with the potential to significantly impact real-world applications.
title Online Distillation with Continual Learning for Cyclic Domain Shifts
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2304.01239