Image Reconstruction on ImageNet 256x256 (test)
38.4PSNR (Reconstruction)Spatial Functa
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Spatial Functa#Param=1.4M‡, Conditioning dimensions=65K2026.06 | 38.4 | — | |
| OursLatent Shape=256x2562023.10 | 37.7 | — | |
| FunctaLatent Shape=32x32x642023.10 | 37.7 | — | |
| SpatialLatent Shape=16x16x2562023.10 | 37.2 | — | |
| Spatial Functa (comparable latent budget)Conditioning dimensions=16K2026.06 | 28.3 | — | |
| LH-NeF#Param=1.8M2026.06 | 28.3 | — | |
| ENF#Param=817K‡2026.06 | 27.5 | — | |
| LDMLatent Shape=64x64x32023.10 | 27.4 | — | |
| DALL-EEmbedding size and |Z|=32, 81922021.06 | — | 34.3 | |
| VQ-HiTEmbedding size and |Z|=16, 10242021.06 | — | 6.37 | |
| VQ-VAEEmbedding size and |Z|=32, 10242021.06 | — | 75.19 | |
| VQ-VAE-2Embedding size and |Z|=64, 512 and 32, 5122021.06 | — | 10 | |
| VQGANEmbedding size and |Z|=16, 10242021.06 | — | 8 |