UDL
UDL is a practicable framework used in Deep Learning (computer vision).
Benchmark
codes, results and models are available in UDL, please contact @Liang-Jian Deng (corresponding author)
Pansharpening model zoo:
- PNN (RS'2016)
- PanNet (CVPR'2017)
- DiCNN1 (JSTAR'2019)
- FusionNet (TGRS'2020)
- DCFNet (ICCV'2021)
Results of DCFNet
Quantitative results
wv3 | SAM | ERGAS |
---|---|---|
new_data10 | 3.934 | 2.531 |
new_data11 | 4.133 | 2.630 |
new_data12_512 | 4.108 | 2.712 |
new_data6 | 2.638 | 1.461 |
new_data7 | 3.866 | 2.820 |
new_data8 | 3.257 | 2.210 |
new_data9 | 4.154 | 2.718 |
Avg(std) | 3.727(0.571) | 2.440(0.474) |
Ideal Value | 0 | 0 |
wv3_1258 | SAM | ERGAS |
---|---|---|
Avg(std) | 3.377(1.200) | 2.257(0.910) |
Ideal Value | 0 | 0 |
Visual results
please see the paper and the sub-directory: ./UDL/results/DCFNet
Install [Option]
please run python setup.py develop
Usage
open UDL/panshaprening/tests, run the following code:
python run_DCFNet.py
Note that default configures don't fit other environments, you can modify configures in pansharpening/models/DCFNet/option_DCFNet.py.
Benefit from mmcv/config.py, the project has the global configures in Basis/option.py, option_DCFNet inherits directly from Basis/option.py.
1. Data preparation
You need to download WorldView-3 datasets.
The directory tree should be look like this:
|-$ROOT/datasets
├── pansharpening
│ ├── training_data
│ │ ├── train_wv3_10000.h5
│ │ ├── valid_wv3_10000.h5
│ ├── test_data
│ │ ├── WV3_Simu
│ │ │ ├── new_data6.mat
│ │ │ ├── new_data7.mat
│ │ │ ├── ...
│ │ ├── WV3_Simu_mulExm
│ │ │ ├── test1_mulExm1258.mat
2. Training
args.eval = False, args.dataset='wv3'
3. Inference
args.eval = True, args.dataset='wv3_singleMat'
Plannings
Please expect more tasks and models
-
pansharpening
- models
-
derain
- models
-
HISR
- models
Contribution
We appreciate all contributions to improve UDL. Looking forward to your contribution to UDL.
Citation
If you use this toolbox or benchmark in your research, please cite this project.
@InProceedings{Wu_2021_ICCV,
author = {Wu, Xiao and Huang, Ting-Zhu and Deng, Liang-Jian and Zhang, Tian-Jing},
title = {Dynamic Cross Feature Fusion for Remote Sensing Pansharpening},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2021},
pages = {14687-14696}
}
Acknowledgement
- MMCV: OpenMMLab foundational library for computer vision.
- HRNet : High-resolution networks and Segmentation Transformer for Semantic Segmentation
License & Copyright
This project is open sourced under GNU General Public License v3.0