Table 2 Best performing hyperparameters for EfficientNetV2-S, MobileNetV3-L, ResNet101, and Swin V2-B according to the bayesian optimization and the hyperband algorithm (BOHB) algorithm.
From: Identification of plant-parasitic nematode genera in turfgrass using deep learning algorithms
Model | Learning rate | Dropout rate (%) | Maximum random rotation (°) | Maximum random brightness | Size of classification head FC layer | Batch size | Best epoch |
|---|---|---|---|---|---|---|---|
EfficientNetV2-S | 7.37 × 10− 5 | 4.61 | 12.77 | 0.0631 | 495 | 32 | 71 |
MobileNetV3-L | 1.68 × 10− 4 | 25.97 | 18.09 | 0.0986 | 674 | 32 | 40 |
ResNet101 | 2.20 × 10− 5 | 17.71 | 16.57 | 0.0317 | 700 | 32 | 50 |
Swin V2-B | 1.77 × 10− 4 | 25.09 | 12.44 | 0.0944 | 921 | 64 | 20 |
