Classification on Clevr-Count VTAB-1K (test)
86.4AccuracyLoRA_mul+VPT_add + PACE
Evaluation Results
| Method | Links | |
|---|---|---|
| LoRA_mul+VPT_add + PACEPre-training Method=MAE, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 86.4 | |
| LoRA_mul+VPT_addPre-training Method=MAE, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 82.1 | |
| LoRA_mul+VPT_add + PACEPre-training Method=DINO, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 61 | |
| LinearPre-training Method=MAE, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 57.1 | |
| LoRA_mul+VPT_addPre-training Method=DINO, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 55.7 | |
| FullPre-training Method=MAE, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 52.5 | |
| LinearPre-training Method=DINO, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 44.2 | |
| FullPre-training Method=DINO, Backbone=ViT-B/16, Pre-trained on=ImageNet-1K2024.09 | 34.5 |