Image-to-Text Retrieval on Flickr30K (I2T metric)
92.9I2T ScoreCOSMOS
Evaluation results
| Method | Links | |
|---|---|---|
| COSMOSData Size=30M, Batch Size=4k, Zero-shot=true2024.12 | 92.9 | |
| MobileCLIP-BData Size=1B, Batch Size=65k, Zero-shot=true2024.12 | 91.4 | |
| VeCLIPData Size=300M, Batch Size=32k, Zero-shot=true2024.12 | 91.2 |