In this software, we provided the following implementations in PyTorch:
- Model and Dataset implementation (
model.pyanddataset.py): U-Net based model (1x1 resolution) for precipitation nowcasting and estimation, and the dataset for each target task. - Codes for experiments: Codes for pre-training and fine-tuning the model, and evaluating the performance of the trained model.
- Authors: Jihoon Ko*, Kyuhan Lee*, Hyunjin Hwang*, Seok-Geun Oh, Seok-Woo Son, Kijung Shin
- Version: 1.0
In order to download example checkpoints for running demos, use the following commands:
$ wget https://www.dropbox.com/s/oitxstngis1ht1d/pretraining_10.pkt -O precipitation\ nowcasting/checkpoints/pretraining_10.pkt
$ wget https://www.dropbox.com/s/p3062yite6lvnq3/finetuned.pkt -O precipitation\ nowcasting/example_checkpoints/finetuned.pktFor more details, see user_guide.pdf.