Our project is based on Nalbach et al. 2017 paper. In this project, a set of buffers are provided to a Deep Convolutional Network in order to synthetize differend shading effects (such as Ambient Occlusion, Depth of Field, Global Illumination and Sub-surface Scattering). The set of buffers depends os the shading effect we want to synthetize.

Input Bufers:

Result (Ambient Occlusion):

0. Python 3.x
1. Tensorflow 1.10
2. Keras
3. OpenCV 3.4(for loading,resizing images)
4. h5py(for saving trained model)
5. pyexr

1. Clone the repo
   
2. Download the dataset (https://deep-shading-datasets.mpi-inf.mpg.de/)
   
3. Install the requirements
   
4. Generate the .tfrecord for trainning and validation (Use the DataReader.py)
   
5. Run "Shading.py"

[1] Oliver Nalbach, Elena Arabadzhiyska, Dushyant Mehta, Hans-Peter Seidel, Tobias Ritschel Deep Shading: Convolutional Neural Networks for Screen-Space Shading to appear in Proc. EGSR 2017