Widescale Weather Witchcraft is a GAN-based application designed to augment road images to simulate various weather conditions. This project leverages deep learning to generate realistic transformations of clear weather road scenes into conditions such as rain, fog, and snow.
A well documented report of our work can be accessed here - Report For Widescale Weather Witchcraft
Click the image below to watch our project explanation video:
You Can access more results inside BestResults directory for translation of weather Conditions and directory ImageSegmentation for results on the segmentation tasks.
 |
 |
 |
 |
Figure 1: View of the Home/Landing page for the application. Users can drag-and-drop images or videos here.
Figure 2: Image showing the loading bar on the app, indicating how much time is remaining until all images are processed.
Figure 3: Snapshot of the app's interface after images have been processed, showcasing the results.
Figure 4: Users can view some of their previously generated transformations from the last 10 minutes.
- Windows Operating System
- Python 3.10
- GPU with CUDA support
Ensure that Python 3.10 is installed on your system. It is required for compatibility with TensorFlow's GPU functionalities on Windows. You can download it from Python's official website.
Clone this repository to your local machine. Use the following command in your terminal:
git clone https://github.com/MasterKinjalk/Gan-App-For-Data-Generation.git
cd Gan-App-For-Data-GenerationIt's recommended to use a virtual environment to manage dependencies:
python -m venv venv
venv\Scripts\activateInstall all required Python packages from the requirements.txt file:
pip install -r requirements.txtIt is also adviced to install torch using torch's -Start Local Website This ensures you have correct CUDA and cuDNN version setup.
To use TensorFlow with GPU support, ensure that you have the correct version of CUDA and cuDNN installed. As of TensorFlow 2.10.0 (compatible with Python 3.10), CUDA 11.2 and cuDNN 8.1 are required. You can download them from NVIDIA's official site:
Make sure to set up the environment variables correctly as per the installation guides provided by NVIDIA.
To make sure that you can replicate our results, we have made our weights available.
-Weights Please read the provided Readme.txt file carefully and unzip the given weights in their correct directories.
To start the application, run the tkinterApp.py script after ensuring all settings are configured correctly:
python tkinterApp.pyContributions to Widescale Weather Witchcraft are welcome! Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.
This project is licensed under the MIT License - see the LICENSE file for details.
Here are some useful resources and references that have influenced our project:
- CycleGAN and pix2pix GitHub repository.
- AU-GAN GitHub Repository
- Adversarial Unsupervised Representation Learning for Activity Time-Series Data (Research Paper)
- Foggy CycleGAN GitHub Repository
- Berkeley DeepDrive Dataset
- Summer to Winter Yosemite Dataset on Kaggle
- Cityscapes Dataset
- FoggySynscapes GitHub Repository
- RESIDE Standard Dehaze Datasets
- DeSnowNet: Context-Aware Deep Network for Snow Removal (Research)
- Time of Day Dataset on Kaggle
- Project Explanation Video on YouTube
This list includes both academic papers and practical resources that provide a foundation and tools for further exploration and understanding of our project's underlying technologies.