Image Reconstruction on COCO 5k 2017 (val)
3.25rFIDSSDD-XL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SSDD-XL#D=153.8M, Resolution=256x2562025.10 | 3.25 | — | — | 0.05 | |
| SSDD-L#D=85.2M, Resolution=256x2562025.10 | 3.29 | — | — | 0.052 | |
| SSDD-M#D=48.0M, Resolution=256x2562025.10 | 3.41 | — | — | 0.054 | |
| SSDD-B#D=20.2M, Resolution=256x2562025.10 | 3.62 | — | — | 0.057 | |
| ε-VAE (H)#D=355.6M, Resolution=256x2562025.10 | 3.65 | — | — | — | |
| SSDD-S#D=13.4M, Resolution=256x2562025.10 | 3.77 | — | — | 0.059 | |
| ε-VAE (M)#D=49.3M, Resolution=256x2562025.10 | 3.98 | — | — | — | |
| SDXL-VAEratio=8, dim=4, training_data=Unknown, resolution=256x2562024.06 | 4.07 | 25.76 | 0.845 | — | |
| LlamaGenratio=8, dim=8, size=16384, training_data=ImageNet, resolution=256x2562024.06 | 4.19 | 24.2 | 0.822 | — | |
| SD-VAEratio=8, dim=4, training_data=Unknown, resolution=256x2562024.06 | 4.45 | 25.41 | 0.831 | — | |
| Consistency decoder#D=625.1M, Resolution=256x2562025.10 | 4.53 | — | — | 0.063 | |
| OAI-Decoderratio=8, dim=4, training_data=Unknown, resolution=256x2562024.06 | 4.59 | 24.19 | 0.8 | — | |
| KL-VAE#D=47.2M, Resolution=256x2562025.10 | 4.65 | — | — | 0.063 | |
| VQGANratio=8, dim=4, size=16384, training_data=OpenImage, resolution=256x2562024.06 | 5.89 | 23.08 | 0.771 | — | |
| DiTo-XL + LPIPS#D=592.2M, Resolution=256x2562025.10 | 5.95 | — | — | 0.076 | |
| VQGANratio=8, dim=4, size=256, training_data=OpenImage, resolution=256x2562024.06 | 6.58 | 22.28 | 0.744 | — | |
| LlamaGenratio=16, dim=8, size=16384, training_data=ImageNet, resolution=256x2562024.06 | 8.11 | 20.42 | 0.678 | — | |
| VQGANratio=16, dim=256, size=16384, training_data=ImageNet, resolution=256x2562024.06 | 12.29 | 19.57 | 0.63 | — | |
| VQGANratio=16, dim=256, size=1024, training_data=ImageNet, resolution=256x2562024.06 | 16.95 | 19.08 | 0.613 | — |