Image to Text (I2T) on Oogiri-GO (test)
39.83T1Qwen-VL+CLoT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen-VL+CLoTSize=7B2023.12 | 39.8 | 35.1 | 22.7 | 64.4 | 40.5 | |
| Qwen-VL+AITSize=7B2023.12 | 38.8 | 30.5 | 15.7 | 62.3 | 36.8 | |
| LLaVA-1.5Size=13B2023.12 | 29.3 | 22.7 | 3.9 | 60.9 | 29.2 | |
| GPT4v2023.12 | 29.1 | 15.1 | 3.9 | 60.4 | 27.1 | |
| Qwen-VLSize=7B2023.12 | 23.2 | 23.1 | 11.9 | 62.2 | 30.1 | |
| mPLUG-Owl MultilingualSize=7B2023.12 | 19.2 | 18.6 | 6 | 60.5 | 26.1 | |
| VisualGLM-6BSize=6B2023.12 | 14.3 | 20.4 | 8.8 | 61.9 | 26.4 | |
| MiniGPT-v2Size=7B2023.12 | 5.3 | 4 | 3.8 | 60.5 | 18.4 |