Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Optimizing Rank-based Metrics with Blackbox Differentiation

About

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. The code is available at https://github.com/martius-lab/blackbox-backprop

Michal Rol\'inek, V\'it Musil, Anselm Paulus, Marin Vlastelica, Claudio Michaelis, Georg Martius• 2019

Related benchmarks

TaskDatasetResultRank
Object DetectionPASCAL VOC 2007 (test)--
821
Image RetrievalCUB-200-2011 (test)
Recall@164
251
Image RetrievalStanford Online Products (test)
Recall@178.6
220
In-shop clothes retrievalin-shop clothes retrieval dataset (test)
Recall@188.1
78
Image RetrievalSOP (test)
Recall@178.6
42
Image RetrievalStanford Online Products (SOP) standard (test)
Recall@178.6
27
Image RetrievaliNaturalist (test)
Recall@162.9
24
Image RetrievalCars196 standard (test)
Recall@184.2
23
Image ClassificationCIFAR-10 (test)
AUCPR0.944
11
Deep Metric LearningiNaturalist
R@152.3
8
Showing 10 of 11 rows

Other info

Follow for update