Alternating Back-Propagation for Generator Network

Tian Han* , Yang Lu* , Song-Chun Zhu , and Ying Nian Wu
* Equal contributions.

University of California, Los Angeles (UCLA), USA

Main > Indirect Face

Recovery error

We learn the model from the compressively sensed data (Can-dès, Romberg, and Tao, 2006). We generate a set of white noise images as random projections. We then project the training images on these white noise images. We can then learn the model from the random projections instead of the original images. We show the recovery error for different latent dimension d, where the recovery error is defined as the per pixel difference between the original image and the recovered image.

experiment d = 20 d = 60 d = 100
error .0795 .0617 .0625

Original images

Recovery images with d = 20

Recovery images with d = 60

Recovery images with d = 100