Skip to content

malvela/WindTurbine-IceDetection_GenerativeAI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Wind Turbine Rotor Blade Generalised Ice Detection with Generative AI

Supplementary material for the application paper "Domain-Invariant Icing Detection on Wind Turbine Rotor Blades with Generative AI for Deep Transfer Learning" (in submission).

A repository with code for predicting blade icing on images of turbine rotor blades using supervised (neural style transfer) and unsupervised (CycleGAN) techniques.

Download and use of the repository:

To download this repository and its submodules use

git clone --recurse-submodules https://github.com/malvela/WindTurbine-IceDetection_GenerativeAI.git

Individual files and functionality:

This repository contains Python files for generalised icing prediction (domain-invariant - independent of the wind park the AI model has been trained on) on wind turbine rotor blades using a tiny computer.

  • CycleGAN/cyclegan_generativeai_icing.py : Used to train the CycleGAN model from scratch (or leverage the pre-trained Summer2Winter Yosemite model).
  • Fast_Style_Transfer/Overlay_Images.ipynb : Used to overlay the styled image to the rotorblade using labelled masks.
  • Fast_Style_Transfer/StyleTransfer_Notebook_BladeImages.ipynb: Used to modify the content images with the pretrained style transfer model.
  • StyleTransfer-TrainFromScratch/NST_TrainingFromScratch.py: Used to modify content images by training neural style transfer algorithm (based on VGG-19) from scratch.

Cite as:

If you are using this repository in your research, please cite it as:

Chatterjee J., Alvela Nieto M.T., Gelbhardt H., Dethlefs N., Ohlendorf J.-H., Greulich A., Thoben K.-D., "Domain-Invariant Icing Detection on Wind Turbine Rotor Blades with Generative AI for Deep Transfer Learning" (in submission)

References:

  1. Reference for CycleGAN original model (and original pre-trained models used - such as summer2winter): (https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix).
  2. Reference for Fast Style Transfer technique: https://www.tensorflow.org/tutorials/generative/style_transfer
  3. Reference for Neural Style Transfer Algorithm: https://www.tensorflow.org/tutorials/generative/style_transfer

License:

This repo is based on the MIT License, which allows free use of the provided resources, subject to the original sources being credit/acknowledged appropriately. The software/resources under MIT license is provided as is, without any liability or warranty at the end of the authors.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published