Image-to-Text on ImageScript (test)
24.5Top-1 R-PrecisionUniPose
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| UniPoseTraining Strategy=Joint training on all tasks2024.11 | 24.5 | 35.4 | 43.2 | 18.2 | 42.5 | 44.7 | |
| UniPoseTraining Strategy=Training on single task2024.11 | 22.4 | 32.8 | 41.2 | 18.2 | 42.4 | 45.2 | |
| GPT4VTraining Strategy=Zero-shot or baseline2024.11 | 17.7 | 24 | 32.3 | 7.1 | 29.1 | 34.2 | |
| Qwen-VLTraining Strategy=Zero-shot or baseline2024.11 | 8.9 | 15.6 | 19.8 | 1.4 | 15.9 | 21.6 | |
| LLaVATraining Strategy=Zero-shot or baseline2024.11 | 5.7 | 12 | 18.9 | 3.2 | 21.8 | 32.9 |