Revisiting Training Strategies and Generalization Performance in Deep Metric Learning

Karsten Roth, Timo Milbich, Samarth Sinha, Prateek Gupta, Bjorn Ommer, Joseph Paul Cohen
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:8242-8252, 2020.

Abstract

Deep Metric Learning (DML) is arguably one of the most influential lines of research for learning visual similarities with many proposed approaches every year. Although the field benefits from the rapid progress, the divergence in training protocols, architectures, and parameter choices make an unbiased comparison difficult. To provide a consistent reference point, we revisit the most widely used DML objective functions and conduct a study of the crucial parameter choices as well as the commonly neglected mini-batch sampling process. Under consistent comparison, DML objectives show much higher saturation than indicated by literature. Further based on our analysis, we uncover a correlation between the embedding space density and compression to the generalization performance of DML models. Exploiting these insights, we propose a simple, yet effective, training regularization to reliably boost the performance of ranking-based DML models on various standard benchmark datasets. Code and a publicly accessible WandB-repo are available at https://github.com/Confusezius/Revisiting_Deep_Metric_Learning_PyTorch.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-roth20a, title = {Revisiting Training Strategies and Generalization Performance in Deep Metric Learning}, author = {Roth, Karsten and Milbich, Timo and Sinha, Samarth and Gupta, Prateek and Ommer, Bjorn and Cohen, Joseph Paul}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {8242--8252}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/roth20a/roth20a.pdf}, url = {https://proceedings.mlr.press/v119/roth20a.html}, abstract = {Deep Metric Learning (DML) is arguably one of the most influential lines of research for learning visual similarities with many proposed approaches every year. Although the field benefits from the rapid progress, the divergence in training protocols, architectures, and parameter choices make an unbiased comparison difficult. To provide a consistent reference point, we revisit the most widely used DML objective functions and conduct a study of the crucial parameter choices as well as the commonly neglected mini-batch sampling process. Under consistent comparison, DML objectives show much higher saturation than indicated by literature. Further based on our analysis, we uncover a correlation between the embedding space density and compression to the generalization performance of DML models. Exploiting these insights, we propose a simple, yet effective, training regularization to reliably boost the performance of ranking-based DML models on various standard benchmark datasets. Code and a publicly accessible WandB-repo are available at https://github.com/Confusezius/Revisiting_Deep_Metric_Learning_PyTorch.} }
Endnote
%0 Conference Paper %T Revisiting Training Strategies and Generalization Performance in Deep Metric Learning %A Karsten Roth %A Timo Milbich %A Samarth Sinha %A Prateek Gupta %A Bjorn Ommer %A Joseph Paul Cohen %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-roth20a %I PMLR %P 8242--8252 %U https://proceedings.mlr.press/v119/roth20a.html %V 119 %X Deep Metric Learning (DML) is arguably one of the most influential lines of research for learning visual similarities with many proposed approaches every year. Although the field benefits from the rapid progress, the divergence in training protocols, architectures, and parameter choices make an unbiased comparison difficult. To provide a consistent reference point, we revisit the most widely used DML objective functions and conduct a study of the crucial parameter choices as well as the commonly neglected mini-batch sampling process. Under consistent comparison, DML objectives show much higher saturation than indicated by literature. Further based on our analysis, we uncover a correlation between the embedding space density and compression to the generalization performance of DML models. Exploiting these insights, we propose a simple, yet effective, training regularization to reliably boost the performance of ranking-based DML models on various standard benchmark datasets. Code and a publicly accessible WandB-repo are available at https://github.com/Confusezius/Revisiting_Deep_Metric_Learning_PyTorch.
APA
Roth, K., Milbich, T., Sinha, S., Gupta, P., Ommer, B. & Cohen, J.P.. (2020). Revisiting Training Strategies and Generalization Performance in Deep Metric Learning. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:8242-8252 Available from https://proceedings.mlr.press/v119/roth20a.html.

Related Material