Training OOM of CPU

[KMnO4AndWaston]

Hello author! Very Nice Job! I try to use Neural-Recon-Wild to reconstruct a 200mx200m region on street, but code will dump every time when it come to update octree. Can you give me some suggestion on it？In my config, Level is set to 5, voxel size is caluculated to 4 meter, scene like below:

[Burningdust21]

Hi, what exactly do you mean by "code will dump every time when it come to update octree". If the CPU is OOM, you can try lowering image resolution or increasing fine-level voxel size. It would be helpful to provide more information, like the error message, and the number of training images used.