Table 2 Details of hyperparameters.
From: Land use classification using multi-year Sentinel-2 images with deep learning ensemble network
Method | Batch size | Trainable parameter | Learning rate | Optimizer | Loss | Momentum | Threshold |
|---|---|---|---|---|---|---|---|
Fully Convolutional Networks | 8 | 134, 270, 278 | 1 × 10−4 | Stochastic gradient descent | Cross-entropy | – | – |
High-Resolution Net | 8 | 9,524,036 | 1 × 10−4 | Adam Optimizer | Dice loss | – | – |
DeepLabv3 + | 8 | 39,756,963 ResNet50 | 1 × 10−2 | Stochastic gradient descent | Cross-entropy | – | – |
UNet | 16 | 14,326,275 | 1 × 10−5 | Nadam | BCE | 0.9 | 0.5 |
ResUNet | 16 | 1,048,953 | 1 × 10−5 | Nadam | BCE | 0.9 | 0.5 |
IRUNet (Proposed) | 16 | 28,864,481 | 1 × 10−5 | Nadam | BCE | 0.9 | 0.5 |
