Please use this identifier to cite or link to this item:
|Publication type:||Article in scientific journal|
|Type of review:||Peer review (publication)|
|Title:||Is it enough to optimize CNN architectures on ImageNet?|
|Published in:||Frontiers in Computer Science|
|Publisher / Ed. Institution:||Frontiers Research Foundation|
|Subjects:||Deep learing; CNN architecture design; ImageNet; Empirical study|
|Subject (DDC):||006: Special computer methods|
|Abstract:||Classification performance based on ImageNet is the de-facto standard metric for CNN development. In this work we challenge the notion that CNN architecture design solely based on ImageNet leads to generally effective convolutional neural network (CNN) architectures that perform well on a diverse set of datasets and application domains. To this end, we investigate and ultimately improve ImageNet as a basis for deriving such architectures. We conduct an extensive empirical study for which we train 500 CNN architectures, sampled from the broad AnyNetX design space, on ImageNet as well as 8 additional well known image classification benchmark datasets from a diverse array of application domains. We observe that the performances of the architectures are highly dataset dependent. Some datasets even exhibit a negative error correlation with ImageNet across all architectures. We show how to significantly increase these correlations by utilizing ImageNet subsets restricted to fewer classes. These contributions can have a profound impact on the way we design future CNN architectures and help alleviate the tilt we see currently in our community with respect to over-reliance on one dataset.|
|Fulltext version:||Published version|
|License (according to publishing contract):||CC BY 4.0: Attribution 4.0 International|
|Departement:||School of Engineering|
|Organisational Unit:||Centre for Artificial Intelligence (CAI)|
|Published as part of the ZHAW project:||RealScore – Scanning of Real-World Sheet Music for a Digital Music Stand|
|Appears in collections:||Publikationen School of Engineering|
Files in This Item:
|2022_Tuggener-Schmidhuber-Stadelmann_Optimize-CNN-ImageNet_fcomp.pdf||7.98 MB||Adobe PDF|
Show full item record
Tuggener, L., Schmidhuber, J., & Stadelmann, T. (2022). Is it enough to optimize CNN architectures on ImageNet? Frontiers in Computer Science, 4(1041703). https://doi.org/10.3389/fcomp.2022.1041703
Tuggener, L., Schmidhuber, J. and Stadelmann, T. (2022) ‘Is it enough to optimize CNN architectures on ImageNet?’, Frontiers in Computer Science, 4(1041703). Available at: https://doi.org/10.3389/fcomp.2022.1041703.
L. Tuggener, J. Schmidhuber, and T. Stadelmann, “Is it enough to optimize CNN architectures on ImageNet?,” Frontiers in Computer Science, vol. 4, no. 1041703, 2022, doi: 10.3389/fcomp.2022.1041703.
Tuggener, Lukas, et al. “Is It Enough to Optimize CNN Architectures on ImageNet?” Frontiers in Computer Science, vol. 4, no. 1041703, 2022, https://doi.org/10.3389/fcomp.2022.1041703.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.