Lately I've been trying to catch up with the latest developments in light-fields. My first project has been trying to understand Stereo Magnification: Learning View Synthesis using Multiplane Images from Google (2018). Now I've created a Colab that uses PyTorch and trains a model at low-res in less than 15 minutes using a small subset of the Real Estate 10k dataset. I also use DeepView's viewer (inside the Colab) to show the results.