Lens-Flare-Reduction-by-Flare-Simulation

Lens Flare Reduction by Flare Simulation and U-Net

1. Prepare Image Set (Image Set Used in Training: https://www.kaggle.com/arnaud58/landscape-pictures)
2. Apply Flare Simulation to Image Set using flare_simulation.py
3. Create Dataloader using dataloader.py
4. Run train.py to start training
5. Run predict.py to predict flare area and inpaint flared image

• .ipynb file is for Colab environment
• Final model available in models
• Training structure inspired by: https://towardsdatascience.com/creating-and-training-a-u-net-model-with-pytorch-for-2d-3d-semantic-segmentation-dataset-fb1f7f80fe55