Improving the Efficiency-Accuracy Trade-Off of DETR-Style Models in Practice
Abstract
We aim to provide a comprehensive view of the inference efficiency of DETR-style detection models. We explore the effect of basic efficiency techniques and identify the factors that are easy to implement, yet effectively improve the efficiency-accuracy trade-off. Specifically, we investigate the effect of input resolution, multi-scale feature enhancement, and backbone pre-training. Our experiments support that 1) adjusting the input resolution is a simple yet effective way to achieve a better efficiency-accuracy trade-off. 2) Multi-scale feature enhancement can be lightened with a marginal decrease in accuracy, and 3) improved backbone pre-training can further improve the trade-off.
Cite
Text
Suh et al. "Improving the Efficiency-Accuracy Trade-Off of DETR-Style Models in Practice." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2024. doi:10.1109/CVPRW63382.2024.00801Markdown
[Suh et al. "Improving the Efficiency-Accuracy Trade-Off of DETR-Style Models in Practice." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2024.](https://mlanthology.org/cvprw/2024/suh2024cvprw-improving/) doi:10.1109/CVPRW63382.2024.00801BibTeX
@inproceedings{suh2024cvprw-improving,
title = {{Improving the Efficiency-Accuracy Trade-Off of DETR-Style Models in Practice}},
author = {Suh, Yumin and Kim, Dongwan and Aich, Abhishek and Schulter, Samuel and Su, Jong-Chyi and Han, Bohyung and Chandraker, Manmohan},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2024},
pages = {8027-8031},
doi = {10.1109/CVPRW63382.2024.00801},
url = {https://mlanthology.org/cvprw/2024/suh2024cvprw-improving/}
}