Style Transfer on CLIP-based evaluation set
78.92AccuracyInST
Evaluation results
| Method | Links | ||
|---|---|---|---|
| InSTAttention module=Multi-layer cross-attention, Dropout=True2022.11 | 78.92 | 80.72 | |
| InSTAttention module=Multi-layer cross-attention, Dropout=False2022.11 | 72.25 | 66.53 | |
| InSTAttention module=MLP, Dropout=True2022.11 | 71.03 | 66.99 | |
| Textual Inversion2022.11 | 68.84 | 69.03 | |
| SDM2022.11 | 63.83 | — |