IEEE CVPR MAT 2024

Best Paper Award

Test-time Specialization of Dynamic Neural Networks 

Abstract:

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.



Pre-camera PDF 

IEEE Library Access

@INPROCEEDINGS{Leroux:MAT2024, 
author={Leroux, Sam and Katare, Dewant and Ding, Aaron Yi and Simoens, Pieter}, 
booktitle=IEEE CVPR 1st Workshop on Test-Time Adaptation: Model, Adapt Thyself (MAT)}, 
title={Test-time Specialization of Dynamic Neural Networks},
year={2024}
}

Sam Leroux, Dewant Katare, Aaron Yi Ding, Pieter Simoens, "Test-time Specialization of Dynamic Neural Networks", in Proceedings of IEEE CVPR 1st Workshop on Test-Time Adaptation: Model, Adapt Thyself (MAT), 2024.