Hi everyone,
I am currently working on a large-scale aerial mapping project and processing a dataset of roughly three thousand high-resolution drone images in Pix4Dmapper. While the initial processing step completes without any issues, I am hitting a severe system lockup whenever the software enters the step two point cloud densification phase.
The primary problem is that as soon as the point cloud generation reaches about forty percent, my workstation’s RAM usage spikes to absolute capacity, causing Pix4Dmapper to suddenly crash to the desktop without writing any crash logs. This issue sets off a cascade of secondary frustrations: my local temp directory gets cluttered with corrupted cache files, the active project file occasionally locks up, and I have to manually clear out the processing folder before I can even attempt to restart the run. I am trying to determine if the local GPU driver is failing to handle the parallel processing threads under such a heavy load, or if the software is simply running out of virtual memory because of how aggressively it caches intermediate spatial data during deep densification loops.
I have been testing a modular system resource tracker that I grabbed as a free download [external link removed by a moderator] to see if limiting the active thread allocation or forcing the system to flush the temporary GPU buffers at regular intervals keeps the process from crashing my machine. Unfortunately, even when restricting the software to only use half of my available processor cores, the system still encounters deep processing delays and eventually hangs during the final point extraction. I am trying to figure out if it is better practice to manually partition my massive image dataset into three separate sub-projects and merge the resulting point clouds afterward, or if there is a specific configuration in the processing options that I can adjust to keep the local memory allocation stable.
If anyone else here has managed dense mapping projects or resolved unexpected desktop crashes during heavy point cloud generation loops, how do you tune your workstation resources? I would be incredibly grateful for any advice or hardware configuration tips to help get this project processed smoothly.