For the final project of the multimedia information retrieval class in my CS Master's I'm doing something similar. The goal is to predict depth maps from singl

For the final project of the multimedia information retrieval class in my CS Master's I'm doing something similar. The goal is to predict depth maps from single-view pictures using CNNs. But how to gather input, output pairs to train them? I'm using a mod for Minecraft I wrote to generate such labeled data, e.g. input: https://i.imgur.com/JRtmCeG.png https://i.imgur.com/JRtmCeG.png output: https://i.imgur.com/q4KocXn.png https://i.imgur.com/q4KocXn.png. I'm not done yet so I can't tell you the results, but it's a fun project.