Accident-causing road structure prediction on BDD100K (val)
86.31RecallQwen2-VL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2-VLZero-shot=true, Trained on=Mapillary2026.03 | 86.31 | 24.81 | 37.09 | 24.36 | |
| Long-CLIPVision Encoder=Long-CLIP, Text Encoder=Long-CLIP, Zero-shot=true, Trained on=Mapillary2026.03 | 83.14 | 90.07 | 86.46 | 71.45 | |
| CLIPVision Encoder=CLIP, Text Encoder=Long-CLIP, Zero-shot=true, Trained on=Mapillary2026.03 | 81.7 | 90.86 | 86.04 | 71.15 | |
| ResNet-50Vision Encoder=ResNet-50, Text Encoder=Long-CLIP, Zero-shot=true, Trained on=Mapillary2026.03 | 81.1 | 88.37 | 84.58 | 68.35 | |
| InternVL2Zero-shot=true, Trained on=Mapillary2026.03 | 76.32 | 26.68 | 37.48 | 25.75 | |
| ViT-BVision Encoder=ViT-B, Text Encoder=Long-CLIP, Zero-shot=true, Trained on=Mapillary2026.03 | 75.12 | 82.14 | 78.47 | 63.02 | |
| LLaVA-1.5Zero-shot=true, Trained on=Mapillary2026.03 | 73.88 | 17.34 | 25 | 16.3 | |
| GPT-4oZero-shot=true, Trained on=Mapillary2026.03 | 51.31 | 19.92 | 27.94 | 18.88 | |
| Phi-3Zero-shot=true, Trained on=Mapillary2026.03 | 41.7 | 22.13 | 27.48 | 17.47 |