Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
chihy 75baa3aebb | 2 years ago | |
---|---|---|
.idea | 2 years ago | |
TDT Dataset | 2 years ago | |
data | 2 years ago | |
dehazing_celery | 2 years ago | |
file_upload | 2 years ago | |
log | 2 years ago | |
net | 2 years ago | |
small tools | 2 years ago | |
0251.jpg | 2 years ago | |
Dehazing_celery.py | 2 years ago | |
celery_main.py | 2 years ago | |
dehaze_api_test.py | 2 years ago | |
readme.md | 2 years ago | |
requirements.txt | 2 years ago | |
start_celery_task.sh | 2 years ago | |
state.bak | 2 years ago | |
state.dat | 2 years ago | |
state.dir | 2 years ago | |
去雾 demo.mp4 | 2 years ago | |
去雾+检测 demo.mp4 | 2 years ago |
PCID-Net:Peng Cheng Image Dehazing Network
Cong Wang
针对雾天采集的交通场景图片及视频,引入自动去雾能力,对图片及视频进行增强。
雾霾视频经算法去雾后,在准确率99%条件下,车辆检测召回率提高6.3%。
采用目标检测算法对检测去雾前后图片中的车辆,在准确率99%条件下比较召回率。
训练数据集:RESIDE-OTS 室外交通场景雾霾数据集 (https://sites.google.com/view/reside-dehaze-datasets/reside-%CE%B2).
测试数据集:TDT Dataset (trafic dehaze test dataset) 自建交通场景雾霾测试数据集。
代码运行的环境与依赖。如下所示:
类别 | 名称 | 版本 |
---|---|---|
os | ubuntu | 16.04 |
计算框架 | NVIDIA GPU + CUDA | 10.0 |
语言 | python | 3.6.8 |
深度学习框架 | pytorch | 1.0.0 |
包依赖 | numpy | 1.19.1 |
matplotlib | 3.3.0 | |
tensorboardX | 2.1 |
快速配置
在本 README 所在的文件夹下,直接执行以下命令配置环境
pip install -r requirements.txt
代码的输入与输出。如下所示:
名称 | 说明 |
---|---|
输入 | 带雾RGB图像。大小为224X224(宽x高) |
输出 | 去雾RGB图像。大小为224X224(宽x高) |
/data
文件夹中存放有来自于 OTS
数据集的24对训练图片和8对验证图片,供验证训练代码的可用性
在 OTS
数据集上训练模型,将文件路径切换至 ./net
下,可直接执行 train.sh
cd ./net
sh train.sh
训练得到的模型权重已放入 net/trained_models/
文件夹中
测试图片放在 test_imgs/
文件夹中
测试在 OTS
数据集上训练的模型,将文件路径切换至 ./net
下,可直接执行 test.sh
程序运行结束后,可在 pred_CID_ots/
文件夹中得到处理后的图片。
cd ./net
sh test.sh
注意:如果待处理文件是视频,首先需要采用 ./small tools
文件夹下的 shipin_chaifen.m
文件将待处理视频拆分为逐帧图片,采用 PCID-Net 对逐帧图片去雾后,再使用 ./small tools
文件夹下的 shipin_pinjie.m
文件将逐帧去雾图片拼接成去雾视频。
./small tools
文件夹下的文件运行环境为 MATLAB 2019b。
针对交通道路雾天采集的交通场景图片及视频,引入自动去雾能力,对图片及视频进行增强
Python MATLAB Shell Text Markdown
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》