Table 5 CNN Architecture (18 layers, learning rate: 0.0001, optimizer: Adam53, epoch: 20).
From: Near Real-Time Wildfire Progression Monitoring with Sentinel-1 SAR Time Series and Deep Learning
Layer No. | Operation | Output | |
|---|---|---|---|
Train | Test | ||
0 (input) | — | 7 × 55 × 55 | 7 × (w + 54) × (h + 54) |
1 | 20@3 × 3Conv + ReLU | 20 × 53 × 53 | 20 × (w + 52) × (h + 52) |
2–8 | 20@7 × 7Conv + ReLU | 20 × 17 × 17 | 20 × (w + 16) × (h + 16) |
9–17 | 20@3 × 3Conv + ReLU | 20 × 1 × 1 | 20 × w × h |
18 | 2@1 × 1Conv + Sigmoid | 2 × 1 × 1 | 2 × w × h |
