We introduce a new architecture called a conditional invertible neural network (cINN), and use it to address the task of diverse image-to-image translation for natural images.
Recent work demonstrated that flow-based invertible neural networks are promising tools for solving ambiguous inverse problems.
Mixture Density Networks are a tried and tested tool for modelling conditional probability distributions.
We demonstrate these properties for the tasks of MNIST digit generation and image colorization.
Many recent invertible neural architectures are based on coupling block designs where variables are divided in two subsets which serve as inputs of an easily invertible (usually affine) triangular transformation.
no code implementations • 8 Mar 2019 • Tim J. Adler, Lynton Ardizzone, Anant Vemuri, Leonardo Ayala, Janek Gröhl, Thomas Kirchner, Sebastian Wirkert, Jakob Kruse, Carsten Rother, Ullrich Köthe, Lena Maier-Hein
Assessment of the specific hardware used in conjunction with such algorithms, however, has not properly addressed the possibility that the problem may be ill-posed.
Often, the forward process from parameter- to measurement-space is a well-defined function, whereas the inverse problem is ambiguous: one measurement may map to multiple different sets of parameters.