3D Object Detection on KITTI Cyclist (val)
72.25AP (Moderate)Full Fine-tune
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Full Fine-tuneTrainable Params=8.65M2025.12 | 72.25 | 93.81 | 68.23 | |
| From ScratchTrainable Params=8.65M2025.12 | 70.91 | 92.68 | 66.65 | |
| SOP2Trainable Params=0.82M2025.12 | 68.24 | 86.69 | 64.56 | |
| SOP2+LoRATrainable Params=0.84M2025.12 | 67.9 | 84.18 | 63.98 | |
| LoRATrainable Params=0.47M2025.12 | 66.96 | 86.82 | 63.42 | |
| BitFitTrainable Params=0.47M2025.12 | 66.67 | 86 | 63.49 | |
| Prompt GeneratorTrainable Params=1.64M2025.12 | 66.44 | 83.71 | 63.1 | |
| Prompt TokenTrainable Params=0.45M2025.12 | 64.92 | 83.6 | 61.5 | |
| Head Fine-tuneTrainable Params=0.45M2025.12 | 63.57 | 82.57 | 60.47 | |
| PL: F-PointNet*Modality=Stereo, Pre-training=extra Scene Flow dataset2020.01 | 25.2 | 41.3 | 24.9 | |
| DSGNModality=Stereo2020.01 | 24.27 | 37.87 | 23.15 | |
| M3D-RPNModality=Mono2020.01 | 2.81 | — | — |