This repository contains various state-of-the-art deep learning models designed to mitigate turbulence effects in videos and images. These models enhance visual clarity and stability, making them suitable for applications in surveillance, remote sensing, and scientific imaging.
- Video Turbulence Mitigation: Deep learning models specifically designed to reduce turbulence in video sequences.
- Image Turbulence Mitigation: Specialized models for enhancing single images affected by turbulence.
- Pre-trained Models: Access to pre-trained models for immediate use.
- Customizable: Easily fine-tune models for specific use cases and datasets.
- Comprehensive Documentation: Detailed guides on installation, usage, and model customization.
Each model has its own dependencies and should be run in separate virtual environments. Refer to the individual model's README file for specific installation instructions.
Refer to the individual model's README file for usage instructions and examples.
Description: This is a Single Frame Atmospheric Turbulence Mitigation: A Benchmark Study and A New Physics-Inspired Transformer Model.
- Folder:
models/TurbNet - Pre-trained Model: Download here
- Documentation: TurbNet Documentation
Description: This model learns to Restore Images Degraded by Atmospheric Turbulence Using Uncertainty, recognized as Best paper at IEEE International Conference on Image Processing, 2021.
- Folder:
models/AT_Net - Pre-trained Model: Download here
- Documentation: AT_Net Documentation
Description: This state-of-the-art transformer-based model mitigates turbulence effects in both videos and images.
- Folder:
models/TMT - Pre-trained Model: Download here
- Documentation: TMT Documentation