I recently installed WebODM and I am trying to process 140 images. My first question. Does it matter if the images are georeferenced? Will this cut down processing time if you are working with georeferenced images? I tried running the fast ortho option on these but it failed saying that I should try increasing the min-num features. I increased this to 10,000. My dataset contains a lot of trees so maybe this is the problem? My last question is related to the CPU. I noticed that when WebODM is processing data my CPU is only running at 15%. Can I allow it to utilize more of my CPU? I am currently running this on an I7 4 core processor.
Yes, we can do preemptive matching by looking at image pairs that are within a certain geographical boundary instead of matching all of them brute-force style, so it will be faster.
Likely, trees are notoriously difficult to reconstruct, try increasing it to 20,000.
If you are running WebODM on Windows using docker, check your docker settings. You might be able to allocate more CPUs to the virtual machine that runs the docker environment. For the rest, some parts of ODM are parallelized, some aren’t, but there’s nothing on your part to tweak, we try to use all available resources whenever possible (which doesn’t mean we can’t improve in certain areas, quite the contrary, there are many places where we could improve).