Ask questions about the questions in the paper

[BloodLemonS]

It is mentioned in the paper that the dynamic mask and dynamic stream are updated. The dynamic mask was initialized when the network structure was initialized. I would like to ask how the dynamic mask was initialized. How exactly did you choose? In the follow-up issue of SLAM, is it necessary for the system to provide the initial dynamic mask to carry out the subsequent dual-stream display？？