NVAE w/ flow

Reported on 4 benchmarks across 1 task · 1 paper · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Medical4 results

Image GenerationonFFHQ 256 x 256
bits/dimension· 2020-07-08
0.69
best: 0.53 (Efficient-VDVAE)
SOTA
NVAE: A Deep Hierarchical Variational Autoencoder arXiv:2007.03898
Image GenerationonImageNet 32x32
bpd· 2020-07-08
3.92
best: 3.34 (NFDM)
NVAE: A Deep Hierarchical Variational Autoencoder arXiv:2007.03898
Image GenerationonCelebA 256x256
bpd· 2020-07-08
0.7
best: 0.51 (Efficient-VDVAE)
NVAE: A Deep Hierarchical Variational Autoencoder arXiv:2007.03898
Image GenerationonCIFAR-10
FID· 2020-07-08
32.53
best: 1.22 (GMem)
NVAE: A Deep Hierarchical Variational Autoencoder arXiv:2007.03898