A video diffusion model with two-stage training improves tropical cyclone image forecasts over a frame-by-frame diffusion baseline and extends the reliable forecast horizon from 36 to 50 hours.
Cyclone track forecasting based on satellite images using artificial neural networks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Improving Tropical Cyclone Forecasting With Video Diffusion Models
A video diffusion model with two-stage training improves tropical cyclone image forecasts over a frame-by-frame diffusion baseline and extends the reliable forecast horizon from 36 to 50 hours.