Skip to content

mhn2836/dehaze

Repository files navigation

My dehaze algorithm

This is the PyTorch implementation for my work based on CVPR'21 paper 4KDehazing. The model can removal hazy, smoke.

Setup:

依赖的库

torch, numpy, tqdm, torchvision, kornia, opencv-python

Training:

将带雾训练数据集放在./haze 文件夹下 对应的清晰数据集放在./gt文件夹下。

运行命令 python train.py。

训练过程可在./result文件夹下找到。

模型保存在./model文件夹下。

Test model:

提供自己基于部分OTS数据集训练的模型dehaze.pth。

将需要测试的数据集放在./test文件下。

运行命令 python test_model.py。

自行建立test_res文件夹,测试结果可在./test_res文件夹下找到。

SSIM_PSNR Test:

ssim_psnr.py

Time inference:

time.py

Other work

https://github.com/mhn2836/YOLOv5s-joint

Cite

Cite: { title = {Ultra-High-Definition Image Dehazing via Multi-Guided Bilateral Learning}, booktitle = {CVPR}, year = {2021} }

About

dehaze algorithm work

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages