Vehicle dimension estimation on VMMGR internal
0.2589Absolute Length ErrorGemini Pro 2.5
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Gemini Pro 2.5N (# of context images)=102026.05 | 0.2589 | 0.0767 | 0.0752 | 5.49 | 3.7 | 4.05 | 88.31 | 99.26 | |
| Gemini Flash 2.5N (# of context images)=102026.05 | 0.261 | 0.0786 | 0.0764 | 5.54 | 3.79 | 4.12 | 88.13 | 99.63 | |
| Gemini Pro 2.5N (# of context images)=32026.05 | 0.2635 | 0.0778 | 0.0765 | 5.6 | 3.76 | 4.12 | 88.08 | 98.82 | |
| Gemini Flash 2.5N (# of context images)=32026.05 | 0.2678 | 0.0805 | 0.0784 | 5.67 | 3.88 | 4.22 | 87.94 | 99.03 | |
| Pixtral LargeN (# of context images)=32026.05 | 0.2837 | 0.0909 | 0.095 | 6.02 | 4.38 | 5.15 | 86.35 | 96.91 | |
| Claude Sonnet 4N (# of context images)=102026.05 | 0.2911 | 0.0853 | 0.0933 | 6.19 | 4.11 | 5.01 | 86.63 | 97.36 | |
| Llama 4 MaverickN (# of context images)=32026.05 | 0.2946 | 0.0957 | 0.1025 | 6.22 | 4.62 | 5.56 | 85.59 | 86.92 | |
| BaselineN (# of context images)=-2026.05 | 0.301 | 0.1157 | 0.0939 | 6.34 | 5.57 | 4.94 | 84.97 | 100 | |
| Claude Sonnet 4N (# of context images)=32026.05 | 0.3142 | 0.0885 | 0.0995 | 6.68 | 4.27 | 5.37 | 85.81 | 93.1 |