Intelligent Systems

Optimizing Rank-based Metrics with Blackbox Differentiation

2020

Conference Paper

al


Rank-based metrics are some of the most widely used criteria for performance evaluation of computer vision models. Despite years of effort, direct optimization for these metrics remains a challenge due to their non-differentiable and non-decomposable nature. We present an efficient, theoretically sound, and general method for differentiating rank-based metrics with mini-batch gradient descent. In addition, we address optimization instability and sparsity of the supervision signal that both arise from using rank-based metrics as optimization targets. Resulting losses based on recall and Average Precision are applied to image retrieval and object detection tasks. We obtain performance that is competitive with state-of-the-art on standard image retrieval datasets and consistently improve performance of near state-of-the-art object detectors.

Author(s): Michal Rolínek and Vít Musil and Anselm Paulus and Marin Vlastelica and Claudio Michaelis and Georg Martius
Book Title: Proceedings of the IEEE/CVF Conerence on Computer Vision and Pattern Recognition (CVPR)
Pages: 7620-7630
Year: 2020
Month: June
Day: 13-19
Publisher: IEEE

Department(s): Autonomous Learning
Research Project(s): Combinatorial Optimization as a Layer / Blackbox Differentiation
Bibtex Type: Conference Paper (inproceedings)
Paper Type: Conference

DOI: 10.1109/CVPR42600.2020.00764
Event Name: IEEE International Conference on Computer Vision and Pattern Recognition (CVPR) 2020
Event Place: Seattle, WA, USA

Digital: True
ISBN: 978-1-7281-7168-5
Note: Best paper nomination
State: Published
Talk Type: Oral
URL: https://openaccess.thecvf.com/content_CVPR_2020/html/Rolinek_Optimizing_Rank-Based_Metrics_With_Blackbox_Differentiation_CVPR_2020_paper.html

Links: Paper @ CVPR2020
Long Oral
Short Oral
Arxiv
Code
Pdf
Video:
Video:

BibTex

@inproceedings{Rolinek2020optimizing,
  title = {Optimizing Rank-based Metrics with Blackbox Differentiation},
  author = {Rolínek, Michal and Musil, Vít and Paulus, Anselm and Vlastelica, Marin and Michaelis, Claudio and Martius, Georg},
  booktitle = {Proceedings of the IEEE/CVF Conerence on Computer Vision and Pattern Recognition (CVPR)},
  pages = {7620-7630},
  publisher = {IEEE},
  month = jun,
  year = {2020},
  note = {Best paper nomination},
  doi = {10.1109/CVPR42600.2020.00764},
  url = {https://openaccess.thecvf.com/content_CVPR_2020/html/Rolinek_Optimizing_Rank-Based_Metrics_With_Blackbox_Differentiation_CVPR_2020_paper.html},
  month_numeric = {6}
}