Image Generation on FFHQ unconditional
2.61FIDVQGAN-LC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VQGAN-LC# Tokens=256, Codebook Size=100,000, Underlying generative model=GPT (404M) [2]2024.06 | 2.61 | 99.1 | |
| ViT-VQGAN (GPT-650M)# Tokens=1,024, Codebook Size=8,192, Underlying generative model=GPT-650M2024.06 | 3.13 | — | |
| VQGAN-FC# Tokens=256, Codebook Size=16,384, Underlying generative model=GPT (404M) [2]2024.06 | 3.23 | 10.9 | |
| VQGAN-EMA# Tokens=256, Codebook Size=16,384, Underlying generative model=GPT (404M) [2]2024.06 | 4.87 | 68.2 | |
| DQTransformer (GPT-308M)# Tokens=640, Codebook Size=1,024, Underlying generative model=GPT-308M2024.06 | 4.91 | — | |
| LDM# Tokens=4,096, Codebook Size=8,1922024.06 | 4.98 | — | |
| Stackformer (GPT-651M)# Tokens=1,024, Codebook Size=1,024, Underlying generative model=GPT-651M2024.06 | 5.67 | — | |
| Stackformer (GPT-307M)# Tokens=1,024, Codebook Size=1,024, Underlying generative model=GPT-307M2024.06 | 6.84 | — | |
| Stackformer (GPT-307M)# Tokens=256, Codebook Size=1,024, Underlying generative model=GPT-307M2024.06 | 7.67 | — | |
| VQGAN-LC# Tokens=256, Codebook Size=100,000, Underlying generative model=LDM [3]2024.06 | 12.3 | 99.7 | |
| VQGAN-EMA# Tokens=256, Codebook Size=16,384, Underlying generative model=LDM [3]2024.06 | 12.5 | 68.2 | |
| VQGAN-FC# Tokens=256, Codebook Size=16,384, Underlying generative model=LDM [3]2024.06 | 13.2 | 11.2 |