SpaSE-UNet3D: Sensor-Driven Wildfire Detection and Progression Prediction from VIIRS Multispectral Imagery
A systematic label-quality audit reveals that many fires lack ground-truth annotations, and a spatial squeeze-and-excitation 3D U-Net whose spatial-only (1,3,3) convolutions avoid temporal mixing on short observation windows, while SE channel attention reweights the VIIRS spectral bands dynamically is proposed.