Computer Vision News - April 2021

The paper opens up to huge opportunities in the field. Transforming images acquired in one axis to acquisitions in another axis could be a very powerful resource; this is taken a step further by performing specific tasks on both scans, even when ground truth segmentations are not available in one of the domains. If you are not yet fascinated by this work, you should definitely watch the video below, another great example of transformation and task prediction which fully shows the excellent and novel work conducted by this group. This can also be found on their github page which I highly suggest visiting if interested. 11 What does Michael Black wish Computer Vision News for our 5th anniversary? Find out on page 38! Unsupervised Domain Adaptation from ...