PiCNet: Physics-infused Convolution Network for Radar-Based Precipitation Nowcasting The PiCNet is a pytorch-based model for precipitation nowcasting and the structure of PiCNet is in PiCNet-structure.
For more information or papers, please refer to PiCNet.
tool.py: This file contains some preprocessing function, such as data transfer function, evaluate function, show picture function, etc.
PiCNet/PiCNet.py: This file is the kernel file, it builds the whole model, contains the advection simulator module and the physics-guided prediction module and refining network module.
Adveciton/test.py & train.py： The former contains the test process of the Advection smutilator. The train.py contains the training process of the Advection smutilator.
PiCNet/test.py & train.py: The former contains the test process of the model. The train.py contains the training process of the model.
Firstly you should apply for the KNMI dataset, you can apply for the dataset by KNMI.
Then, you can use PiCNet/Adveciton/train.py and PiCNet/PiCNet/train.py to train your new model.
The results of the Advection smutilator are trained by PiCNet/Adveciton/train.py.
You can use PiCNet/PiCNet/test.py to test your model.
Python 3.6+, Pytorch 1.0 and Ubuntu.