tjulitianyi 64358c56f2 | 2 years ago | |
---|---|---|
checkpoint | 2 years ago | |
figures | 2 years ago | |
test | 2 years ago | |
README.md | 2 years ago | |
adjustment_net_train.py | 2 years ago | |
calculate_ssim_psnr.py | 2 years ago | |
decomposition_net_train.py | 2 years ago | |
evaluate.py | 2 years ago | |
evaluate_LOLdataset.py | 2 years ago | |
model.py | 2 years ago | |
reflectance_restoration_net_train.py | 2 years ago | |
train_testcase.sh | 2 years ago | |
utils.py | 2 years ago |
KinD的Tensorflow实现
Kindling the Darkness: a Practical Low-light Image Enhancer. In ACMMM2019
Yonghua Zhang, Jiawan Zhang, Xiaojie Guo
PSNR | SSIM | |
---|---|---|
比赛精度基线 | 20 | 0.75 |
Paper精度基线 | 20.8665 | 0.8022 |
Ours(Ascend 910) | 21.7097 | 0.8864 |
|-- checkpoint ----模型存放路径
|--decom_net_train ----自动生成的模型存放路径
|--illumination_adjust_net_train ----自动生成的模型存放路径
|--Restoration_net_train ----自动生成的模型存放路径
|-- figures ----展示图片存放路径
|-- LOLdataset ----数据集存放路径(由下载的数据集解压得到)
|-- test ----示例图片存放文件夹
|-- adjustment_net_train.py ----网络模型文件
|-- calculate_ssim_psnr.py ----验收指标计算文件
|-- decomposition_net_train.py ----网络模型文件
|-- evaluate_LOLdataset.py ----LOLdataset数据集评估文件
|-- evaluate.py ----评估文件
|-- model.py ----网络模型文件
|-- README.md ----使用前必读
|-- reflectance_restoration_net_train.py ----网络模型文件
|-- train_testcase.sh ----NPU训练启动shell
|-- utils.py ----其他作用文件
在本级目录下,执行如下命令:
0. 下载数据集
首先,下载数据集压缩包:数据集可以从这里下载 BaiduNetdisk or google drive(该数据集压缩包将LOL数据集的训练对保存在“./LOLdataset/our485/”下,并将评估对保存在“./LOLdataset/eval15/”下)
其次,将得到的LOLdataset_and_expo.zip放到本级代码目录下,之后执行如下命令解压缩即可
unzip LOLdataset_and_expo.zip
激活环境
source activate /home/ma-user/miniconda3/envs/TensorFlow-1.15.0/
运行脚本,运行时间大概需要30719s,请耐心等待
bash train_testcase.sh
查看结果
cat calculate_ssim_psnr.log
在打印的最后,可以看到我们的结果为
##### Test Average SSIM: 0.8864641895358436 #####
##### Test Average PSNR: 21.709714716708934 #####
如果您想直接进行测试,我们也提供了解决方案。
首先,下载预训练模型 BaiduNetdisk or google drive, 之后运行
python evaluate.py
我们预先训练过的模式已经改变了。因此,结果与本文的报告有一定的差异。但是,可以调整照明率以获得更好的效果。
我们的代码部分参考了 code.
@inproceedings{zhang2019kindling,
author = {Zhang, Yonghua and Zhang, Jiawan and Guo, Xiaojie},
title = {Kindling the Darkness: A Practical Low-light Image Enhancer},
booktitle = {Proceedings of the 27th ACM International Conference on Multimedia},
series = {MM '19},
year = {2019},
isbn = {978-1-4503-6889-6},
location = {Nice, France},
pages = {1632--1640},
numpages = {9},
url = {http://doi.acm.org/10.1145/3343031.3350926},
doi = {10.1145/3343031.3350926},
acmid = {3350926},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {image decomposition, image restoration, low light enhancement},
}
第四届中国软件开源创新大赛·赛道二:任务挑战赛(模型王者挑战赛):基于华为Ascend 910,利用Tensorflow 1.15.0 实现KinD CV类图像增强网络,数据集:训练集/测试集:Low Light paired dataset(LOL) 最终精度:PSNR:21.16,SSIM:0.88
Unity3D Asset Python other
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》