Image Reconstruction on MS-COCO 2017 (val)
1.8rFIDMergeTok-BL (VAE)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MergeTok-BL (VAE)Type=1D VAE+VQ, #Tok.=256, #Code=2^14, Res.=2562026.05 | 1.8 | — | 22.3 | 0.7 | |
| MergeTok-BL (VQ)Type=1D VAE+VQ, #Tok.=256, #Code=2^14, Res.=2562026.05 | 2.4 | — | 21.7 | 0.68 | |
| GigaTok-BLType=1D VQ, #Tok.=256, #Code=2^14, Res.=2562026.05 | 2.6 | — | 21.5 | 0.68 | |
| MergeTok-BL (VAE)Type=1D VAE+VQ, #Tok.=256, #Code=2^14, Res.=5122026.05 | 2.9 | — | 22.8 | 0.71 | |
| MergeTok-BL (VQ)Type=1D VAE+VQ, #Tok.=256, #Code=2^14, Res.=5122026.05 | 3.6 | — | 22.1 | 0.69 | |
| GigaTok-SBType=1D VQ, #Tok.=256, #Code=2^14, Res.=2562026.05 | 3.8 | — | 20.9 | 0.66 | |
| UniTokRatio=16, Resolution=256x2562026.05 | 4.05 | — | 23.92 | 0.73 | |
| GigaTok-BLType=1D VQ, #Tok.=256, #Code=2^14, Res.=5122026.05 | 4.2 | — | 21.9 | 0.69 | |
| WinTokRatio=16, Resolution=256x2562026.05 | 4.24 | — | 23.83 | 0.72 | |
| WeTokResolution=Original Resolution, Ratio=162026.02 | 5.3 | — | 21.94 | 0.59 | |
| SD-VAE 1.xResolution=Original Resolution, Ratio=82026.02 | 5.94 | — | 21.68 | 0.64 | |
| UniWeTokResolution=Resize 256 x 256, Ratio=322026.02 | 6.18 | — | 22.58 | 0.66 | |
| UniWeTokResolution=Original Resolution, Ratio=322026.02 | 6.46 | — | 22.29 | 0.65 | |
| WeTokResolution=Resize 256 x 256, Ratio=162026.02 | 6.55 | — | 21.99 | 0.63 | |
| WeTokRatio=16, Resolution=256x2562026.05 | 6.55 | — | 21.99 | 0.63 | |
| Open-MAGVIT2-I-PTResolution=Original Resolution, Ratio=162026.02 | 6.65 | — | 21.61 | 0.57 | |
| Open-MAGVIT2Ratio=16, Codebook Size=262144, Zero-shot=true2024.12 | 6.76 | — | 22.31 | 0.65 | |
| IBQRatio=16, Codebook Size=262144, Zero-shot=true2024.12 | 6.79 | — | 22.28 | 0.65 | |
| CosmosResolution=Original Resolution, Ratio=162026.02 | 7.23 | — | 20.45 | 0.53 | |
| IBQRatio=16, Codebook Size=16384, Zero-shot=true2024.12 | 7.67 | — | 21.58 | 0.62 | |
| Open-MAGVIT2Ratio=16, Codebook Size=16384, Zero-shot=true2024.12 | 7.93 | — | 22.21 | 0.62 | |
| Open-MAGVIT2-I-PTResolution=Resize 256 x 256, Ratio=162026.02 | 7.93 | — | 22.21 | 0.62 | |
| Open-MAGVIT2-I-PTRatio=16, Resolution=256x2562026.05 | 7.93 | — | 22.21 | 0.62 | |
| LlamaGenRatio=16, Codebook Size=16384, Zero-shot=true, Initial Training=Pre-trained on ImageNet2024.12 | 8.4 | — | 20.28 | 0.55 | |
| LlamaGenResolution=Resize 256 x 256, Ratio=162026.02 | 8.4 | — | 20.28 | 0.55 | |
| LlamaGenRatio=16, Resolution=256x2562026.05 | 8.4 | — | 20.28 | 0.55 | |
| WeTokResolution=Resize 256 x 256, Ratio=322026.02 | 8.94 | — | 20.31 | 0.55 | |
| WeTokResolution=Original Resolution, Ratio=322026.02 | 8.94 | — | 20.31 | 0.55 | |
| Show-oRatio=16, Codebook Size=8192, Zero-shot=true2024.12 | 9.26 | — | 20.9 | 0.59 | |
| Show-oResolution=Resize 256 x 256, Ratio=162026.02 | 9.26 | — | 20.9 | 0.59 | |
| Show-oRatio=16, Resolution=256x2562026.05 | 9.26 | — | 20.9 | 0.59 | |
| CosmosRatio=16, Codebook Size=64000, Zero-shot=true2024.12 | 11.97 | — | 19.22 | 0.48 | |
| CosmosResolution=Resize 256 x 256, Ratio=162026.02 | 11.97 | — | 19.22 | 0.48 | |
| DDIMConditioning=UC, Steps=502022.11 | — | 0.03 | — | — | |
| DDIMConditioning=UC, Steps=2002022.11 | — | 0.023 | — | — | |
| DDIMConditioning=C, Steps=502022.11 | — | 0.42 | — | — | |
| DDIMConditioning=C, Steps=2002022.11 | — | 0.497 | — | — | |
| EDICTConditioning=UC, Steps=502022.11 | — | 0.015 | — | — | |
| EDICTConditioning=UC, Steps=2002022.11 | — | 0.015 | — | — | |
| EDICTConditioning=C, Steps=502022.11 | — | 0.015 | — | — | |
| EDICTConditioning=C, Steps=2002022.11 | — | 0.015 | — | — | |
| LDM AESteps=502022.11 | — | 0.015 | — | — | |
| LDM AESteps=2002022.11 | — | 0.015 | — | — |