FORECAST-CLSTM: A New Convolutional LSTM Network for Cloudage Nowcasting

by Chao Tan, Xin Feng, Jianwu Long, and Li Geng.

This repository contains the source code and dataset for FORECAST-CLSTM, provided by Chao Tan.

The paper is avaliable for download here. Click here for more details.

---

Dataset

Our SCMD2016 dataset is available for download at TianYiCloud(2.5GB) or BaiduCloud(2.5GB) (extraction code: ssby).
SCMD dataset is a brand new cloudage nowcasting dataset for deep learning research. It contains 20000 grayscale image sequences for training and another 3500 image sequences for testing. You can get the SCMD2016 dataset at any time but only for scientific research. At the same time, please cite our work when you use the SCMD dataset.

The mnist dataset in npz format can be download here.

Prerequisites

• Python 3.5
• PyTorch >= 0.4.0
• opencv 0.4
• PyQt 4
• numpy

Train

1. For moving-mnist training, please download mnist.npz dataset and place it in ./data folder, for cloudage nowcasting training, please download and unzip SCMD2016 dataset and place it in ./data folder.
2. For moving-mnist dataset, run python trainer_mnist.py --model "FORECAST_CLSTM_M" --epochs 100 --train-batch 16 --gpu-ids "0" --checkpoint "checkpoint/forecast_clstm_m" to start training.
3. For scmd2016 dataset, run python trainer_scmd.py --model "FORECAST_CLSTM_S" --epochs 100 --train-batch 16 --gpu-ids "0" --checkpoint "checkpoint/forecast_clstm_s"" to start training.

Citation

@inproceedings{
     title={{FORECAST-CLSTM}: A New Convolutional LSTM network for Cloudage Nowcasting},
     author={Tan, Chao and Feng, Xin and Long, Jianwu and Geng, Li},
     booktitle={VCIP},
     year={2018},
     note={to appear},
}