Their results are clearly 'meh' given the number of inputs, in [0] they use only 2-4 cameras in real-time to achieve the same or better quality (see their supplementary material). In [1] is a video view synthesis generated by 4 cameras.
No comments yet.