De-identification on Base Image Dataset (train)
91.94Name IoULoRA
Evaluation results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LoRAAdaptation Strategy=Parameter-Efficient Fine-Tuning (LoRA)2026.06 | 91.94 | 97.28 | 98.49 | 91.96 | 97.14 | 92.05 | 96.01 | 94.19 | 98.01 | 91.44 | 98.49 | 96.9 | 97.14 | 88.98 | 95.01 | 87.82 | |
| FullAdaptation Strategy=Full Fine-tuning2026.06 | 91.34 | 97.33 | 98.49 | 89.84 | 97.81 | 91.03 | 97.19 | 94.82 | 97.36 | 90.45 | 98.09 | 96.15 | 97.62 | 86.26 | 94.26 | 89.2 | |
| Presidio2026.06 | 0.85 | — | — | 0.74 | — | 0.67 | — | 1.19 | — | 0.72 | — | 1.41 | — | 0.74 | — | 0.67 |