Reflection_removal_rendering

Code for
Single Image Reflection Removal with Physically-Based Training Images (CVPR 2020 oral)
Soomin Kim, Yuchi Huo, and Sung-Eui Yoon

https://sgvr.kaist.ac.kr/~smkim/Reflection_removal_rendering/

Please cite this paper if you use this code in an academic publication.

@InProceedings{Kim_2020_CVPR,
author = {Kim, Soomin and Huo, Yuchi and Yoon, Sung-Eui},
title = {Single Image Reflection Removal With Physically-Based Training Images},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}

---

This code is based on tensorflow, and has been tested on Ubuntu 16.04 LTS.

Setup

• $ cd Reflection_removal_rendering
• Create a folder called VGG_Model
• Download pre-trained VGG-19 model (imagenet-vgg-verydeep-19) in this page in VGG-VD models category.
• Move the downloaded pre-trained VGG model(imagenet-vgg-verydeep-19.mat) to VGG_Model folder

Testing

• Download pre-trained model
• $ tar -xvzf pre-trained.tar.gz
• Check a newly created folder pre-trained, whether downloaded model files are in that folder.
• Example test images are provided in test_imgs/blended.
• Run python main.py
• Test results are in the Results folder.

If you want to try your own test images, then change input_path (line 301)in main.py. Also, if you don't have ground truth images for test images, then comment out the quality assess part (line 335-336 in main.py).

Acknowledgement

This reflection removal framework is based upon perceptual-reflection-removal (CVPR 2018), which is modified for our proposed structure.