1 paper
Viorica PÄtrÄucean, Xu Owen He, Joseph Heyward +10
We propose a novel block for \emph{causal} video modelling. It relies on a time-space-channel factorisation with dedicated blocks for each dimension: gated linear recurrent units (…