Image-to-Text Generation on CUB
29CLIP I2T ScoreOurs (EBM)
Evaluation Results
| Method | Links | |
|---|---|---|
| Ours (EBM)Sampling=0.08, NFE=31, Parameter Overhead=2M2026.05 | 29 | |
| Ours (Gen)Sampling=0.001, NFE=12026.05 | 28.6 | |
| CMVAESampling=0.001, NFE=12026.05 | 27.2 | |
| MMVAESampling=0.001, NFE=12026.05 | 23.5 |