Image Classification on 11 image datasets CoOp protocol (test)
0.771Top-1 AccuracyCross-Modal Partial Finetuning
Evaluation results
| Method | Links | |
|---|---|---|
| Cross-Modal Partial FinetuningNumber of shots=16, Train speed=<3min2023.01 | 0.771 | |
| Cross-Modal Linear ProbingNumber of shots=16, Train speed=<1min2023.01 | 0.76 | |
| Cross-Modal AdapterNumber of shots=162023.01 | 0.759 | |
| Tip-AdapterNumber of shots=16, Train speed=5min2023.01 | 0.758 | |
| Tip-Adapter†Number of shots=16, Train speed=5min2023.01 | 0.751 | |
| Cross-Modal WiSE-FTNumber of shots=162023.01 | 0.741 | |
| ProGradNumber of shots=16, Train speed=17hr2023.01 | 0.74 | |
| Cross-Modal PromptingNumber of shots=162023.01 | 0.74 | |
| Cross-Modal Partial FinetuningNumber of shots=8, Train speed=<3min2023.01 | 0.736 | |
| CoOpNumber of shots=16, Train speed=14hr2023.01 | 0.734 | |
| Cross-Modal AdapterNumber of shots=82023.01 | 0.734 | |
| Cross-Modal Linear ProbingNumber of shots=8, Train speed=<1min2023.01 | 0.73 | |
| Tip-AdapterNumber of shots=8, Train speed=5min2023.01 | 0.725 | |
| Tip-Adapter†Number of shots=8, Train speed=5min2023.01 | 0.722 | |
| Cross-Modal WiSE-FTNumber of shots=82023.01 | 0.717 | |
| WiSE-FTNumber of shots=16, Train speed=<1min2023.01 | 0.716 | |
| ProGradNumber of shots=8, Train speed=17hr2023.01 | 0.714 | |
| Cross-Modal PromptingNumber of shots=82023.01 | 0.714 | |
| Linear ProbingNumber of shots=16, Train speed=<1min2023.01 | 0.711 | |
| Cross-Modal AdapterNumber of shots=42023.01 | 0.708 | |
| Cross-Modal Partial FinetuningNumber of shots=4, Train speed=<3min2023.01 | 0.705 | |
| Cross-Modal Linear ProbingNumber of shots=4, Train speed=<1min2023.01 | 0.703 | |
| CoOpNumber of shots=8, Train speed=14hr2023.01 | 0.699 | |
| Tip-AdapterNumber of shots=4, Train speed=5min2023.01 | 0.697 | |
| Tip-Adapter†Number of shots=4, Train speed=5min2023.01 | 0.69 | |
| Cross-Modal WiSE-FTNumber of shots=42023.01 | 0.69 | |
| Cross-Modal PromptingNumber of shots=42023.01 | 0.686 | |
| ProGradNumber of shots=4, Train speed=17hr2023.01 | 0.685 | |
| WiSE-FTNumber of shots=8, Train speed=<1min2023.01 | 0.684 | |
| Cross-Modal AdapterNumber of shots=22023.01 | 0.676 | |
| Cross-Modal Partial FinetuningNumber of shots=2, Train speed=<3min2023.01 | 0.672 | |
| Cross-Modal Linear ProbingNumber of shots=2, Train speed=<1min2023.01 | 0.67 | |
| CoOpNumber of shots=4, Train speed=14hr2023.01 | 0.668 | |
| Tip-AdapterNumber of shots=2, Train speed=5min2023.01 | 0.667 | |
| Cross-Modal WiSE-FTNumber of shots=22023.01 | 0.664 | |
| Tip-Adapter†Number of shots=2, Train speed=5min2023.01 | 0.659 | |
| WiSE-FTNumber of shots=4, Train speed=<1min2023.01 | 0.653 | |
| Linear ProbingNumber of shots=8, Train speed=<1min2023.01 | 0.65 | |
| ProGradNumber of shots=2, Train speed=17hr2023.01 | 0.649 | |
| Cross-Modal PromptingNumber of shots=22023.01 | 0.649 | |
| Cross-Modal Partial FinetuningNumber of shots=1, Train speed=<3min2023.01 | 0.647 | |
| Tip-AdapterNumber of shots=1, Train speed=5min2023.01 | 0.645 | |
| Cross-Modal AdapterNumber of shots=12023.01 | 0.644 | |
| Cross-Modal Linear ProbingNumber of shots=1, Train speed=<1min2023.01 | 0.641 | |
| Cross-Modal WiSE-FTNumber of shots=12023.01 | 0.638 | |
| Tip-Adapter†Number of shots=1, Train speed=5min2023.01 | 0.633 | |
| ProGradNumber of shots=1, Train speed=17hr2023.01 | 0.626 | |
| CoOpNumber of shots=2, Train speed=14hr2023.01 | 0.623 | |
| Cross-Modal PromptingNumber of shots=12023.01 | 0.62 | |
| WiSE-FTNumber of shots=2, Train speed=<1min2023.01 | 0.618 | |
| CoOpNumber of shots=1, Train speed=14hr2023.01 | 0.596 | |
| WiSE-FTNumber of shots=1, Train speed=<1min2023.01 | 0.591 | |
| Zero-Shot CLIPNumber of shots=02023.01 | 0.588 | |
| Linear ProbingNumber of shots=4, Train speed=<1min2023.01 | 0.572 | |
| Linear ProbingNumber of shots=2, Train speed=<1min2023.01 | 0.476 | |
| Linear ProbingNumber of shots=1, Train speed=<1min2023.01 | 0.367 |