Risk Object Detection on DRAMA-Reasoning (val)
61.2mIoUUniDrive
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| UniDriveInput=Video2026.06 | 61.2 | 31 | 66.5 | 83.7 | 60.8 | |
| UniDrive w/o STInput=Image2026.06 | 59.8 | 29.8 | 64.3 | 82.1 | 57.5 | |
| Shikra*Input=Image2026.06 | 50.3 | 10.4 | 59.5 | 73.8 | 50.1 | |
| InstructBLIPInput=Image2026.06 | 47.8 | 9.1 | 62.2 | 74.5 | 48.9 | |
| LLaVAInput=Image2026.06 | 47.2 | 8 | 62.1 | 74.2 | 47.4 | |
| BLIP-2Input=Image2026.06 | 46.3 | 8.1 | 60.2 | 73.7 | 46.2 | |
| eP-ALMInput=Video2026.06 | 43.2 | 7.2 | 56.8 | 68.8 | 47.3 | |
| Video-LLAMAInput=Video2026.06 | 42.8 | 6.9 | 55.3 | 67.9 | 48.4 |