Image Reconstruction on ImageNet 1024x1024 (val)
1.87rFIDMergeTok-BL (VQ)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MergeTok-BL (VQ)Type=1D VAE+VQ, Ratio=16, #Tok.=256, #Code=2^142026.05 | 1.87 | 20.14 | 67 | |
| TiTok-S-128Type=1D VQ, Ratio=32, #Tok.=128, #Code=2^122026.05 | 2.32 | 16.97 | 51 | |
| Layton-HType=1D Diff., Ratio=32, #Tok.=-, #Code=-2026.05 | 2.78 | 19.8 | 65 | |
| LlamaGen-Tok.Type=2D VQ, Ratio=32, #Tok.=8^2, #Code=2^142026.05 | 3.17 | 19.94 | 64 | |
| OpenMAGVIT2Type=2D LFQ, Ratio=16, #Tok.=16^2, #Code=2^182026.05 | 3.43 | 19.45 | 63 | |
| Taming-VQGANType=2D VQ, Ratio=16, #Tok.=16^2, #Code=2^102026.05 | 6.02 | 19.2 | 60 |