D-Softmax splits softmax into independent intra- and inter-class losses, matching ArcFace's face verification accuracy while enabling 1/64 negative-class sampling for faster large-scale training.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Softmax Dissection: Towards Understanding Intra- and Inter-class Objective for Embedding Learning
D-Softmax splits softmax into independent intra- and inter-class losses, matching ArcFace's face verification accuracy while enabling 1/64 negative-class sampling for faster large-scale training.