DiD uses a GAN with a difference encoder that maximizes distances between image-change vectors from different latent axes, reporting improved disentanglement metrics on dSprites and 3DShapes.
Measuring disentanglement: A review of metrics
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Disentanglement in Difference: Directly Learning Semantically Disentangled Representations by Maximizing Inter-Factor Differences
DiD uses a GAN with a difference encoder that maximizes distances between image-change vectors from different latent axes, reporting improved disentanglement metrics on dSprites and 3DShapes.