Image Generation on CIFAR-10 (32x32, latent dim 32, test)
93.53FIDVAE - OURS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VAE - OURSSampling=Geometry-aware2022.09 | 93.53 | 71 | 68 | |
| RAE (SN)Latent Prior=10-component GMM, Regularization=Spectral2022.09 | 114.59 | 32 | 53 | |
| RAELatent Prior=10-component GMM2022.09 | 118.25 | 35 | 57 | |
| RAE (GP)Latent Prior=10-component GMM, Regularization=GP2022.09 | 120.32 | 34 | 58 | |
| RAE (L2)Latent Prior=10-component GMM, Regularization=L22022.09 | 123.25 | 33 | 54 | |
| AE - GMMLatent Prior=10-component GMM2022.09 | 130.28 | 35 | 58 | |
| WAE2022.09 | 132.99 | 24 | 52 | |
| VAE - GMMLatent Prior=10-component GMM2022.09 | 138.25 | 29 | 53 | |
| VAE - N(0, 1)Latent Prior=N(0, 1)2022.09 | 162.58 | 10 | 32 | |
| RHVAE2022.09 | 167.41 | 12 | 22 | |
| AE - N(0, 1)Latent Prior=N(0, 1)2022.09 | 196.5 | 5 | 17 | |
| VAMP2022.09 | 198.14 | 5 | 11 | |
| HVAE2022.09 | 201.7 | 13 | 21 |