Image Classification on SUN397 (val)
77.51AccuracyStelLA
Evaluation Results
| Method | Links | |
|---|---|---|
| StelLAModel=ViT Large, Params (%)=0.542025.10 | 77.51 | |
| DoRAModel=ViT Large, Params (%)=0.552025.10 | 77.19 | |
| PiSSAModel=ViT Large, Params (%)=0.532025.10 | 77.15 | |
| LoRAModel=ViT Large, Params (%)=0.532025.10 | 77.08 | |
| Learnable Memory TokensMemory cells=12022.03 | 77 | |
| Full fine-tuningFine-tuning regime=Full2022.03 | 76.9 | |
| Learnable Memory TokensMemory cells=102022.03 | 76.9 | |
| Learnable Memory TokensMemory cells=202022.03 | 76.9 | |
| Learnable Memory TokensMemory cells=52022.03 | 76.8 | |
| Head + Class token fine-tuningFine-tuning regime=Head + Class2022.03 | 76.4 | |
| StelLAModel=ViT Base, Params (%)=0.732025.10 | 76.28 | |
| Head-only fine-tuningFine-tuning regime=Head Only2022.03 | 75.9 | |
| DoRAModel=ViT Base, Params (%)=0.752025.10 | 75.67 | |
| LoRAModel=ViT Base, Params (%)=0.722025.10 | 75.62 | |
| PiSSAModel=ViT Base, Params (%)=0.722025.10 | 75.41 | |
| Zero-Shot baselineEvaluation Protocol=Zero-Shot, Label Distribution Estimation=None2024.04 | 47.1 | |
| OTTEREvaluation Protocol=Zero-Shot, Label Distribution Estimation=BBSE+OT2024.04 | 25.6 | |
| BBSE+PMEvaluation Protocol=Zero-Shot, Label Distribution Estimation=BBSE+PM2024.04 | 6.9 | |
| Linear Probing baselineEvaluation Protocol=Linear Probing, Label Distribution Estimation=None2024.04 | 0.2 | |
| BBSE+PMEvaluation Protocol=Linear Probing, Label Distribution Estimation=BBSE+PM2024.04 | 0.2 | |
| OTTEREvaluation Protocol=Linear Probing, Label Distribution Estimation=BBSE+OT2024.04 | 0.2 |