Improved Contrast Zero-DCE

• This is the Zero Deep curve estimation method with contrast enhancement and Noise Reduction for enchancing Low Light Images.

• This Project is for non-commercial use only.

• This Project Code is ran on Kaggle GPU 'NVIDIA TESLA P100 GPU'.

Acknowledgements

• Original Zero-DCE Paper by Li-Chongyi

My Additions to the Original Code

• Implemented CLAHE algorithm while training the model.
• Implemented Simplest Colour Balance and Noise Reduction to the results of the model

Requirements

• Python 3.9.7
• PyTorch 1.10.1
• Open CV 4.5.4.60
• torchvision 0.11.2
• Nvidia GPU (I used Kaggle GPU)
• Training Data is taken from the original paper. 432 Images are used for training the model.
• You may use more from this link(this containes 2002 training samples).

Authors

• @Soumyadeep17

Results

Documentation

My Paper for this Code

FAQ

1. Why can't I open the 'Improved_Contrast_Zero_DCE_with_Results.ipynb' file?

The size of the notebook has increased due to inclusion of results. Therefore to view this file you need to download it first to your system.

2. How much time did it take to train the model.

The training took approximately 1.5 hrs on kaggle GPU using 432 images.

Feedback

If you have any feedback, please reach out to me at soumyadeep171198@gmail.com.

If you like it Leave a Star.😊

Thank You for Visiting. 😀

License

MIT