Abstract
We consider image completion from the perspective of amortized inference inan image generative model. We leverage recent state of the art variationalauto-encoder architectures that have been shown to produce photo-realisticnatural images at non-trivial resolutions. Through amortized inference in sucha model we can train neural artifacts that produce diverse, realistic imagecompletions even when the vast majority of an image is missing. We demonstratesuperior sample quality and diversity compared to prior art on the CIFAR-10 andFFHQ-256 datasets. We conclude by describing and demonstrating an applicationthat requires an in-painting model with the capabilities ours exhibits: the useof Bayesian optimal experimental design to select the most informative sequenceof small field of view x-rays for chest pathology detection.