For anyone that wants some images to play with, I have uploaded some Mavic 2 Pro images, 5472 X 3648 pixels. My 4GB VRAM GPU wont extract features at full size, but will if you reduce by 1.5X to 3644 pixels.
All 4 sets are of various sections of a nearby gravel road in poor condition with plenty of potholes and corrugations in places.
GSDs are mostly <1cm, images taken from 25-30metres AGL, mostly off-nadir, along the road and from the sides. Flight was manual with the drone stopped for each image, so no rolling shutter distortion.
DTM
Small part of a point cloud
The 265 image dataset took 19 hours with these settings-
||auto-boundary: true, dem-resolution: 1, dsm: true, dtm: true, feature-quality: ultra, gps-accuracy: 8, matcher-neighbors: 12, mesh-size: 250000, orthophoto-resolution: 0.6, pc-filter: 5, pc-quality: high, pc-rectify: true, resize-to: -1, use-3dmesh: true|
|Average GSD:|0.6 cm| Area:|48,631.13 m²|
116 image dataset took 4h 52m
||auto-boundary: true, dem-resolution: 2, dsm: true, dtm: true, feature-quality: ultra, gps-accuracy: 8, matcher-neighbors: 10, mesh-size: 250000, orthophoto-resolution: 0.6, pc-filter: 5, pc-quality: ultra, pc-rectify: true, resize-to: -1, use-3dmesh: true|
|Average GSD:|0.51 cm| |Area:|1,500.34 m²|
Processed with i7 10700K 3.8GHz CPU, 96GB 2.933GHz RAM
NVIDIA GeForce GTX 1650 Super 4GB Graphics card
https://drive.google.com/drive/folders/1ast12U4qPT2473nVQuLRjntCzDpmicXc?usp=sharing