Adversarial Computer Vision

Adversarial training methods for image-to-image translation & recovery.

Deep neural nets with adversarial training methods for tackling the task of image-to-image translation in its full generality (i.e. with noisy domains), with applications such as style transfer with noisy data, removing image presets in an unsupervised manner.

Project poster (PDF)

James Oldfield, BSc Creative Computing