Residual Flow

Reported on 5 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.

Medical5 results

Image GenerationonMNIST
bits/dimension· 2019-06-06
0.97
best: 0.65 (Locally Masked PixelCNN (8 orders))
SOTA
Residual Flows for Invertible Generative Modeling arXiv:1906.02735
Image GenerationonImageNet 64x64
Bits per dim· 2019-06-06
3.757
best: 4.351 (Logsparse (6 layers))
Residual Flows for Invertible Generative Modeling arXiv:1906.02735
Image GenerationonImageNet 32x32
bpd· 2019-06-06
4.01
best: 3.34 (NFDM)
Residual Flows for Invertible Generative Modeling arXiv:1906.02735
Image GenerationonCelebA 256x256
bpd· 2019-06-06
0.992
best: 0.51 (Efficient-VDVAE)
Residual Flows for Invertible Generative Modeling arXiv:1906.02735
Image GenerationonCIFAR-10
FID· 2019-06-06
46.37
best: 1.22 (GMem)
Residual Flows for Invertible Generative Modeling arXiv:1906.02735