Region-Conditioned Orthogonal 3D U-Net for Weather4Cast Competition
This work addresses weather forecasting accuracy for meteorologists and climate researchers, but it is incremental as it builds on existing 3D U-Net methods with specific modifications.
The paper tackled the problem of predicting super-resolution rain movies from low-resolution satellite data in the Weather4Cast competition by enhancing a 3D U-Net with region-conditioned layers and orthogonality regularizations, achieving up to 19.54% improvement over the baseline with minimal parameter increase.
The Weather4Cast competition (hosted by NeurIPS 2022) required competitors to predict super-resolution rain movies in various regions of Europe when low-resolution satellite contexts covering wider regions are given. In this paper, we show that a general baseline 3D U-Net can be significantly improved with region-conditioned layers as well as orthogonality regularizations on 1x1x1 convolutional layers. Additionally, we facilitate the generalization with a bag of training strategies: mixup data augmentation, self-distillation, and feature-wise linear modulation (FiLM). Presented modifications outperform the baseline algorithms (3D U-Net) by up to 19.54% with less than 1% additional parameters, which won the 4th place in the core test leaderboard.