Table 3 The experiments’ configurations.
From: Cross-dataset late fusion of Camera–LiDAR and radar models for object detection
Configuration | Model | Specification |
|---|---|---|
Dataset splitting | Both | 3-fold stratified cross-validation + 3 seeds (42, 2024, 7) |
Shuffle dataset | Both | Enabled during training |
Optimizer | Both | Adam |
Loss function | Both | Cross-entropy |
Epochs | Both | 10 |
Learning rate | KITTI Model | 1 × 10⁻³ |
Learning rate | Radar Model | 1 × 10⁻⁵ |
Framework | Both | PyTorch, NumPy, Scikit-learn |
Hardware | Both | Google Cloud Platform GPU |
Fusion scheme | Fusion | Weighted fusion (0.6 KITTI, 0.4 Radar) |
Evaluation metrics | Fusion | mAP, per-class AP, mean ± std, 95% CI |
