Hardware-aware mobile building block evaluation for computer vision
- Author
- Maxim Bonnaerens (UGent) , Matthias Freiberger (UGent) , Marian Verhelst and Joni Dambre (UGent)
- Organization
- Project
- Abstract
- In this paper, we propose a methodology to accurately evaluate and compare the performance of efficient neural network building blocks for computer vision in a hardware-aware manner. Our comparison uses pareto fronts based on randomly sampled networks from a design space to capture the underlying accuracy/complexity trade-offs. We show that our approach enables matching of information obtained by previous comparison paradigms, but provides more insights into the relationship between hardware cost and accuracy. We use our methodology to analyze different building blocks and evaluate their performance on a range of embedded hardware platforms. This highlights the importance of benchmarking building blocks as a preselection step in the design process of a neural network. We show that choosing the right building block can speed up inference by up to a factor of two on specific hardware ML accelerators.
- Keywords
- hardware-aware deep learning, edge computing
Downloads
-
applsci-12-12615.pdf
- full text (Published version)
- |
- open access
- |
- |
- 2.37 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-01GMAETX5071Z7MH0DB8BR1NQ9
- MLA
- Bonnaerens, Maxim, et al. “Hardware-Aware Mobile Building Block Evaluation for Computer Vision.” APPLIED SCIENCES-BASEL, vol. 12, no. 24, 2022, doi:10.3390/app122412615.
- APA
- Bonnaerens, M., Freiberger, M., Verhelst, M., & Dambre, J. (2022). Hardware-aware mobile building block evaluation for computer vision. APPLIED SCIENCES-BASEL, 12(24). https://doi.org/10.3390/app122412615
- Chicago author-date
- Bonnaerens, Maxim, Matthias Freiberger, Marian Verhelst, and Joni Dambre. 2022. “Hardware-Aware Mobile Building Block Evaluation for Computer Vision.” APPLIED SCIENCES-BASEL 12 (24). https://doi.org/10.3390/app122412615.
- Chicago author-date (all authors)
- Bonnaerens, Maxim, Matthias Freiberger, Marian Verhelst, and Joni Dambre. 2022. “Hardware-Aware Mobile Building Block Evaluation for Computer Vision.” APPLIED SCIENCES-BASEL 12 (24). doi:10.3390/app122412615.
- Vancouver
- 1.Bonnaerens M, Freiberger M, Verhelst M, Dambre J. Hardware-aware mobile building block evaluation for computer vision. APPLIED SCIENCES-BASEL. 2022;12(24).
- IEEE
- [1]M. Bonnaerens, M. Freiberger, M. Verhelst, and J. Dambre, “Hardware-aware mobile building block evaluation for computer vision,” APPLIED SCIENCES-BASEL, vol. 12, no. 24, 2022.
@article{01GMAETX5071Z7MH0DB8BR1NQ9, abstract = {{In this paper, we propose a methodology to accurately evaluate and compare the performance of efficient neural network building blocks for computer vision in a hardware-aware manner. Our comparison uses pareto fronts based on randomly sampled networks from a design space to capture the underlying accuracy/complexity trade-offs. We show that our approach enables matching of information obtained by previous comparison paradigms, but provides more insights into the relationship between hardware cost and accuracy. We use our methodology to analyze different building blocks and evaluate their performance on a range of embedded hardware platforms. This highlights the importance of benchmarking building blocks as a preselection step in the design process of a neural network. We show that choosing the right building block can speed up inference by up to a factor of two on specific hardware ML accelerators.}}, articleno = {{12615}}, author = {{Bonnaerens, Maxim and Freiberger, Matthias and Verhelst, Marian and Dambre, Joni}}, issn = {{2076-3417}}, journal = {{APPLIED SCIENCES-BASEL}}, keywords = {{hardware-aware deep learning,edge computing}}, language = {{eng}}, number = {{24}}, pages = {{13}}, title = {{Hardware-aware mobile building block evaluation for computer vision}}, url = {{http://doi.org/10.3390/app122412615}}, volume = {{12}}, year = {{2022}}, }
- Altmetric
- View in Altmetric
- Web of Science
- Times cited: