SAR Burn-Scar CNN Detection
Detect burn scars from a 5-channel Sentinel-1 SAR-anomaly raster using a small ONNX CNN (e.g. KTH SAR4Wildfire BAN-2020). The input bands are, in fixed order: 1. VH log-ratio — log10(post / baseline_mean) 2. baseline μ 3. baseline σ 4. baseline p10 5. baseline p90 The model returns a single-channel probability raster, which is thresholded into a binary uint8 burn-scar mask. Optionally the float32 probability raster is returned alongside via `emit_probability: true`.
| ID | cnn_burn_detect |
|---|---|
| Domain | ml |
| Category | inference |
| Type | raster-to-raster |
| Version | 1.0.0 |