May 2019
Intermediate to advanced
272 pages
7h 19m
English
The impactful and amazing work done by Guilin et al. on image inpainting has been featured on both Forbes and Fortune:

In their paper, the authors claim that using regular convolutions for image inpainting led to issues such as color discrepancy and blurriness, and that postprocessing methods to reduce such artifacts are not efficient. Their solution to this problem is through partial convolution, in which the convolution is masked and normalized with respect to the pixels that are valid.
In the visual arts, this work represents a new avenue for exploring and synthesizing content using a smart retouching brush. ...
Read now
Unlock full access