A Model for Single-Image Depth Estimation

This neural net was trained to predict the relative depth map from a single image using a novel technique based on sparse ordinal annotations. Each training example only needs to be annotated with a pair of points and its relative distance to the camera. After training, the net is able to reconstruct the full depth map. See more details here.

This network gives the 2D depth map directly. Brighter points correspond to further regions.

Use ListPlot3D to plot a 3D representation of the depth map.

Combine the original image with the depth plot using the Texture directive.

Or catenate multiple frames to create a smooth animation.