Doppler-angle estimation on Rotation-augmented image corpus Image-level sampling
1.96MAE (°)5-model ensemble (stacked)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| 5-model ensemble (stacked)Hyperparameter Tuning=Optuna-tuned, Aggregation=stacked, Validation=pooled out-of-fold2025.08 | 1.96 | 2.79 | 99.5 | |
| 5-model ensemble (mean)Hyperparameter Tuning=Optuna-tuned, Aggregation=pooled out-of-fold means2025.08 | 2.09 | 3.03 | 99.4 | |
| DenseNet201Hyperparameter Tuning=Optuna-tuned, Validation=5-fold cross-validation2025.08 | 3 | 4.03 | 98.8 | |
| ResNet50Hyperparameter Tuning=Optuna-tuned, Validation=5-fold cross-validation2025.08 | 3.64 | 4.84 | 98.1 | |
| VGG19Hyperparameter Tuning=Optuna-tuned, Validation=5-fold cross-validation2025.08 | 4.22 | 5.79 | 97.6 | |
| InceptionV3Hyperparameter Tuning=Optuna-tuned, Validation=5-fold cross-validation2025.08 | 4.68 | 6.59 | 97 | |
| XceptionHyperparameter Tuning=Optuna-tuned, Validation=5-fold cross-validation2025.08 | 4.75 | 6.59 | 97 |