Zero-shot Classification on ELEVATER ImageNet 21 datasets
56.9Average AccuracyCLIP-FT
Evaluation results
| Method | Links | |
|---|---|---|
| CLIP-FTmode=Fine-tuned2023.05 | 56.9 | |
| CLIP2023.05 | 56.4 | |
| MosaiCLIP WiSE-FTfine_tuning=WiSE-FT2023.05 | 55.9 | |
| Syn-CLIPtraining_data=Million-scale synthetic data2023.05 | 55.3 | |
| MosaiCLIP NoCurriccurriculum_learning=false2023.05 | 54.9 | |
| Teaching SVLCaugmentation=External LLM text augmentation2023.05 | 54.8 | |
| MosaiCLIP2023.05 | 53.5 |