One-dimensional Convolutional Neural Network and Recurrent Neural Network to classify extracted Sentinel-1 pixel-based time-series.
1. Samples of time-seires
2. One-dimensional Convolutional Neural Network
Run the notebook to train 1D CNN Model.
| Weighted precision | Weighted Recall | Weighted F1-score | Accuracy | |
|---|---|---|---|---|
| VV | 0.86 | 0.857 | 0.854 | 0.857 |
| VH | 0.848 | 0.842 | 0.838 | 0.842 |
| Both | 0.848 | 0.842 | 0.838 | 0.842 |
3. Recurrent Neural Network
Run the notebook to train RNN Model.
| Weighted precision | Weighted Recall | Weighted F1-score | Accuracy | |
|---|---|---|---|---|
| 1-layer | 0.859 | 0.856 | 0.853 | 0.856 |
| 2-layer | 0.866 | 0.859 | 0.854 | 0.859 |
| 3-layer | 0.829 | 0.826 | 0.815 | 0.826 |
Please read report for more detail.