ConvLSTM
所属分类:数值算法/人工智能
开发工具:Jupyter Notebook
文件大小:10152KB
下载次数:1
上传日期:2022-06-22 01:38:54
上 传 者:
sh-1993
说明: ConvLSTM。pytorch,,
(ConvLSTM.pytorch,,)
文件列表:
.idea (0, 2021-06-17)
.idea\ConvLSTM.pytorch.iml (398, 2021-06-17)
.idea\inspectionProfiles (0, 2021-06-17)
.idea\inspectionProfiles\profiles_settings.xml (174, 2021-06-17)
.idea\misc.xml (194, 2021-06-17)
.idea\modules.xml (284, 2021-06-17)
.idea\vcs.xml (180, 2021-06-17)
.idea\workspace.xml (4417, 2021-06-17)
.ipynb_checkpoints (0, 2021-06-17)
.ipynb_checkpoints\visualize-checkpoint.ipynb (246230, 2021-06-17)
configs (0, 2021-06-17)
configs\__pycache__ (0, 2021-06-17)
configs\__pycache__\config_3x3_16_3x3_32_3x3_64.cpython-37.pyc (2050, 2021-06-17)
configs\__pycache__\config_3x3_32_3x3_64_3x3_128.cpython-37.pyc (2051, 2021-06-17)
configs\config_3x3_16_3x3_32_3x3_64.py (2481, 2021-06-17)
configs\config_3x3_32_3x3_64_3x3_128.py (2486, 2021-06-17)
datas (0, 2021-06-17)
datas\train-images-idx3-ubyte.gz (9912422, 2021-06-17)
main.py (3200, 2021-06-17)
networks (0, 2021-06-17)
networks\BinaryDiceLoss.py (1334, 2021-06-17)
networks\ConvLSTM.py (6115, 2021-06-17)
networks\CrossEntropyLoss.py (635, 2021-06-17)
networks\__pycache__ (0, 2021-06-17)
networks\__pycache__\BinaryDiceLoss.cpython-37.pyc (1351, 2021-06-17)
networks\__pycache__\ConvLSTM.cpython-37.pyc (5500, 2021-06-17)
networks\__pycache__\CrossEntropyLoss.cpython-37.pyc (1026, 2021-06-17)
output (0, 2021-06-17)
output\cache (0, 2021-06-17)
output\cache\099_00010.png (102911, 2021-06-17)
output\cache\099_00030.png (109192, 2021-06-17)
requirements.txt (77, 2021-06-17)
utils (0, 2021-06-17)
utils\__pycache__ (0, 2021-06-17)
utils\__pycache__\dataset.cpython-37.pyc (3091, 2021-06-17)
utils\__pycache__\functions.cpython-37.pyc (2546, 2021-06-17)
utils\__pycache__\utils.cpython-37.pyc (1150, 2021-06-17)
utils\dataset.py (3454, 2021-06-17)
... ...
# ConvLSTM.pytorch
This repository is an unofficial pytorch implementation of
[Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting](https://arxiv.org/abs/1506.04214v1).
We reimplement the experiments in the paper based on the MovingMNIST dataset, which is followed by [Github](https://github.com/jhhuang96/ConvLSTM-PyTorch).
Detailed understanding is available on my [Blog](https://www.cnblogs.com/CZiFan/p/12630949.html).
## Requirements
- Pytorch>=0.4.0
- CPU or GPU
- Other packages can be installed with the following instruction:
```
pip install requirements.txt
```
## Quick start
Running the code with the following command, and the '--config' parameter represents different network architectures.
```
python main.py --config 3x3_16_3x3_32_3x3_***
```
## Results
| Model | Parameters(M) | Flops(G) | DiceLoss |
|---|---|---|---|
| 3x3_16_3x3_32_3x3_*** | 0.61 | 9.19 | 0.682311 |
| 3x3_32_3x3_***_3x3_128 | 2.45 | 36.35 | 0.665905 |
* Note: In order to reduce the number of parameters and flops,
we did not strictly follow the model architecture in the paper, but modified it into unet style structure.
![result1](output/cache/099_00010.png)
![result2](output/cache/099_00030.png)
## Citation
```
@inproceedings{xingjian2015convolutional,
title={Convolutional LSTM network: A machine learning approach for precipitation nowcasting},
author={Xingjian, SHI and Chen, Zhourong and Wang, Hao and Yeung, Dit-Yan and Wong, Wai-Kin and Woo, Wang-chun},
booktitle={Advances in neural information processing systems},
pages={802--810},
year={2015}
}
```
近期下载者:
相关文件:
收藏者: