A pretrained RGB VAE extended with a zero-initialized alpha channel and alpha-aware losses reconstructs RGBA images with substantially higher PSNR/SSIM than prior transparency methods, using much less training data.
Denoising diffusion probabilistic models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
AlphaVAE: Unified End-to-End RGBA Image Reconstruction and Generation with Alpha-Aware Representation Learning
A pretrained RGB VAE extended with a zero-initialized alpha channel and alpha-aware losses reconstructs RGBA images with substantially higher PSNR/SSIM than prior transparency methods, using much less training data.