A unifying mutual information view of metric learning: cross-entropy vs. pairwise losses

Recently, substantial research efforts in Deep Metric Learning (DML) focused on designing complex pairwise-distance losses, which require convoluted schemes to ease optimization, such as sample mining or pair weighting. The standard cross-entropy loss for classification has been largely overlooked in DML... (read more)

PDF Abstract ECCV 2020 PDF ECCV 2020 Abstract

Results from the Paper


Ranked #2 on Metric Learning on Stanford Online Products (using extra training data)

     Get a GitHub badge
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT BENCHMARK
Metric Learning CARS196 ResNet-50 + Cross-Entropy [email protected] 89.3 # 3
Metric Learning CUB-200-2011 ResNet-50 + Cross-Entropy [email protected] 69.2 # 4
Metric Learning In-Shop ResNet-50 + Cross-Entropy [email protected] 90.6 # 3
Metric Learning Stanford Online Products ResNet-50 + Cross-Entropy [email protected] 81.1 # 2

Methods used in the Paper