Introduction

This is a simulator that generates foggy, rainy, smoky and cloudy image over a clear remote sensing image.

Update

2024.03.22

• Accelerate speed when processing large images.
• Improve continuity of generated fog.

Preparation

Create a Virtual Environment

Run pip install -r requirements.txt.

Generate Foggy, Rainy and Cloudy Images

Just put the images into the folder data/raw.

Generate Smoky Images

Besides prepare images, you should also prepare labels corresponding to every image in YOLO format, and put them into the folder data/labels.

Usage

Generate Foggy Images

Use command python simulate.py --foggy --light --medium --heavy. Parameters light / medium / heavy control the thickness of the fog. Note that you should at least use one parameter, or there will be no output. Output images will be in the folder data/foggy.

This method is modified from FoHIS.

Generate Rainy Images

Use command Python simulate.py --rainy. Output images will be in the folder data/rainy.

Generate Smoky Images

Use command python simulate.py --smoky --light --medium --heavy. The usage of the last three parameters is same to that in generating foggy images. Output images will be in the folder data/smoky.

Generate Cloudy Images

Use command Python simulate.py --cloudy. Output images will be in the folder data/cloudy.

Demonstration

Foggy Simulation

From left to right is original image, image with light fog, image with medium fog and image with heavy fog respectively.

Rainy Simulation

From left to right is original image, image with light rain and image with heavy rain respectively. The heavy rain simulation is not included in the code because it's not quite real.

Smoke Simulation

From left to right is original image, image with light smoke, image with medium smoke and image with heavy smoke respectively.

Cloudy Simulation

From left to right is original image, image with clouds. We demonstrate two examples here.

Cite

If you found this repository useful, please consider citing this repository:

@misc{Runou2023,
  author = {Runou Yang},
  title = {AdverseWeatherSimulation},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/RicardooYoung/AdverseWeatherSimulation}},
}