System Efficiency Analysis on nuScenes
40.5Transmitted Packet Size (kB)GPT-4o
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-4oModel Type=API, Input Type=Language Description, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 40.5 | 7.5245 | 3.0132 | — | — | — | |
| Qwen-2.5VL-7BModel Type=Local, Input Type=Language Description, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 40.5 | 8.6372 | 5.5699 | 20,522 | 15,819 | 16.54 | |
| FleetAgentModel Type=Local, Input Type=Vectorized V2N Messages, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 40.5 | 4.4116 | 1.061 | 1,241 | 17,081 | — | |
| GPT-4oModel Type=API, Input Type=BEV Images, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 4,218.8 | 6.931 | 2.7076 | — | — | — | |
| Qwen-2.5VL-7BModel Type=Local, Input Type=BEV Images, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 4,218.8 | 5.8987 | 1.053 | 2,604 | 15,819 | 2.1 | |
| GPT-4oModel Type=API, Input Type=Raw Images, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 25,312.5 | 5.8783 | 1.516 | — | — | — | |
| Qwen-2.5VL-7BModel Type=Local, Input Type=Raw Images, Hardware=NVIDIA A100-40GB, Batch Size=12026.06 | 25,312.5 | 12.6158 | 1.0635 | 10,558 | 15,819 | 8.51 |