Novel view synthesis of dynamic scenes is becoming important in various applications, including augmented and virtual reality.

To reveal the effects of extant models, we benchmark them against diverse and hard scenes. Additionally, we construct a cross-subject benchmark pre-trained on large-scale datasets to assess generalizable methods.

Recent work has shown that CNN-based depth and ego-motion estimators can be learned using unlabelled monocular videos. However, the performance is limited by unidentified moving objects that violate the underlying static scene assumption in geometric image reconstruction.

Recovering the skeletal shape of an animal from a monocular video is a longstanding challenge.