13839
resolution of all inputs can lose the detailed textural infor-   effective forward warping [36]. While being able to gen-
mation and considerably decrease the interpolation perfor-     erate sharp and clean interpolations, the sequential process
mance, and using a shallower model likewise reduces the       of explicit motion estimation and frame synthesis may in-
average quality. To circumvent this problem, we propose       crease the computational complexity of the model.
a novel dynamic ...