no code implementations • 14 Mar 2023 • Wang Dai, Archontis Politis, Tuomas Virtanen
Specifically, each mask is used to multiply the corresponding channel's 2D representation, and the masked output of all channels are then summed.
