cuda out of memory with a ply aboout 8mb three hundred thousand points use 4090 24gb

nv-tlabs / NKSR

[CVPR 2023 Highlight] Neural Kernel Surface Reconstruction

https://research.nvidia.com/labs/toronto-ai/NKSR

Other

735 stars 43 forks source link

cuda out of memory with a ply aboout 8mb three hundred thousand points use 4090 24gb #65

Open kuaiqushangzixiba opened 7 months ago

kuaiqushangzixiba commented 7 months ago

...why, I see that the waymo example data is 200m, when I run the chunk code in the chunked example, it will be divided into 3000 chunks, wasting half an hour...

kuaiqushangzixiba commented 7 months ago

运行一个自己的ply点云，大概8MB大小，三十万个点，直接用recons_simple，显示显存不足。然后我想使用recons_chunk中的方法加载，但是进度条显示分了10000多块？并且运行一段时间后电脑就卡死了。。。程序被迫中止。

heiwang1997 commented 6 months ago

哈咯，你的点云长啥样呀，把recons_simple里面的detail_level调小一点到0.1试试看呢？