TY - GEN
T1 - Test-time Specialization of Dynamic Neural Networks
AU - Leroux, Sam
AU - Katare, Dewant
AU - Ding, Aaron Yi
AU - Simoens, Pieter
N1 - Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.
PY - 2024
Y1 - 2024
N2 - In recent years, there has been a notable increase in the size of commonly used image classification models. This growth has empowered models to recognize thousands of diverse object types. However, their computational demands pose significant challenges, especially when deploying them on resource-constrained edge devices. In many use cases where a model is deployed on an edge device, only a small subset of the classes will ever be observed by a given model instance. Our proposed test-time specialization of dynamic neural networks allows these models to become faster at recognizing the classes that are observed frequently, while maintaining the ability to recognize all other classes, albeit slightly less efficient. We benchmark our approach on a real-world edge device, obtaining significant speedups compared to the baseline model without test-time adaptation.
AB - In recent years, there has been a notable increase in the size of commonly used image classification models. This growth has empowered models to recognize thousands of diverse object types. However, their computational demands pose significant challenges, especially when deploying them on resource-constrained edge devices. In many use cases where a model is deployed on an edge device, only a small subset of the classes will ever be observed by a given model instance. Our proposed test-time specialization of dynamic neural networks allows these models to become faster at recognizing the classes that are observed frequently, while maintaining the ability to recognize all other classes, albeit slightly less efficient. We benchmark our approach on a real-world edge device, obtaining significant speedups compared to the baseline model without test-time adaptation.
KW - Edge AI
KW - resource efficient deep learning
KW - Test time adaptation
UR - http://www.scopus.com/inward/record.url?scp=85206468358&partnerID=8YFLogxK
U2 - 10.1109/CVPRW63382.2024.00111
DO - 10.1109/CVPRW63382.2024.00111
M3 - Conference contribution
AN - SCOPUS:85206468358
T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
SP - 1048
EP - 1056
BT - Proceedings - 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2024
PB - IEEE
T2 - 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2024
Y2 - 16 June 2024 through 22 June 2024
ER -