I’ve had success using Split + Split-Overlap in the past on large area jobs, but now i’m trying to test it on a smaller job of much higher resolution.
I have a roof, 100 x 50m, with GSD around 0.1~0.2cm. To try get good quality outputs, I’ve had to set ignore-gsd: true and ortho-resolution to 0.4.
I’ve also been messing around with clusters as i have 3 rack servers laying around and thought i’d give it a test. I tried the following settings but after 8 hours it still hadn’t finished (~250 photos!) so i just stopped it-
min-num-features: 10000, texturing-nadir-weight: 32, orthophoto-resolution: .4, dem-resolution: 0.8, mesh-size: 500000, ignore-gsd: true, mesh-octree-depth: 12, split-overlap: 30, split: 75, fast-orthophoto: true, texturing-data-term: area, rerun-from: dataset
I think the issue was that too many sub models were being created because of the small overlap size, but seeing as the roof was only 100x50m, I thought 30m split overlap would be ideal.
Does anyone have any input?