δLTA: Decoupling Camera Sampling from Processing to Avoid Redundant Computations in the Vision Pipeline | Publicación