The boosting time consumption of incremental shine-mapping on seq00 from MaiCity with 700 > pc_count_gpu_limit

[laliwang]

Thanks for your great work in the implicit mapping of large-scale outdoor scenes! When I test your scripts python shine_incre.py ./config/maicity/maicity_incre_replay.yaml on the whole Maicity seq00 sequence, I notice that the time consumption of mapping(without marching_cubes and visualization) greatly boost when I set the every_frame=1 in the configs, which is supposed to be 16s/it on the CPU. Since the incremantal mapping only process frames of limited sliding windows, what is the necessity of restricting pc_count_gpu_limit=500 in your implemention? And is it possible to lift this restriction for more practical and efficient incremental mapping on GPU?

[YuePanEdward]

Thanks for your interest in our project. It's a good catch. Indeed this pc_count_gpu_limit is not needed for the incremental mapping with sliding windows. You can remove this limit for replaying-based incremental mapping mode.