Image Reconstruction on TextOCR
2.3rFIDMGVQ-G4
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MGVQ-G4Downsample ratio=16, Resolution=512x512, Evaluation Protocol=Zero-shot, Sub-codebooks=42025.07 | 2.3 | 27.95 | |
| SD-VAEDownsample ratio=16, Resolution=512x512, Evaluation Protocol=Zero-shot, Tokenizer Type=Continuous2025.07 | 2.38 | 26.3 | |
| OpenMagvit2Downsample ratio=16, Resolution=512x512, Evaluation Protocol=Zero-shot2025.07 | 4.19 | 23.42 | |
| VARDownsample ratio=16, Resolution=512x512, Evaluation Protocol=Zero-shot2025.07 | 4.49 | 23.19 | |
| LlamaGenDownsample ratio=16, Resolution=512x512, Evaluation Protocol=Zero-shot2025.07 | 4.94 | 22.47 | |
| VQGANDownsample ratio=16, Resolution=512x512, Evaluation Protocol=Zero-shot2025.07 | 6.49 | 21.4 |