Non-parametric Depth Estimation for Images from a Single Reference Depth


We present a non-parametric method for estimating depth of a single still image. We start from a single reference image and its corresponding 3-d depth and use an unsupervised neural network to transform the reference depth to represent the target image. In doing so, we attempt to mimic the human vision capability of perceiving the depth of a given image… (More)


5 Figures and Tables