Processing large data set keeps failing

Yeah corridors are memory heavy; try lowering split further?

Thanks I will give it a try
From the log it seems like the gdal cache is set to a maximum of 60GB
Is there a way to adjust that from within WebODM? I got the version without manually installing docker

I tried two more times. First reducing split to 1500 and then 1000 images. Still fails with the same error message

any other ideas?

Hi, it looks like a nice set. Are the images RGB or MS? Are the images taken on the same date / height ?

What is your required resolution? Fast orthophoto is a nice setting to get a quick and fast overview, but imho it's not the best setting to do calculations / analysis on the data.

For 13000 images you need a big(ger) computer I guess https://docs.webodm.org/installation/#hardware-requirements especially if you want to process the images in defaulf or high res.

hi @Aeret

Sorry for poor editing. discord on a mobile browser is a piece of crap. images are jpg RGB. They were taken across multiple days. Most of them at the same height but the first flights was slightly higher (I tried a set without it as well though). I need full resolution since I need to identify individual flowers . since its mostly a flat area I had hoped to get away with fast-ortho. I still wonder if the issue is related to the limited GDAL cache but I don't know how I could adjust that. Unfortunately it also doesn't seem to save the intermediate steps so I could just try to run that gdal command separately

Thank you for the offer.
I have been able to process subsets (each flight by itself with up to around 800 images) without many issues. But with the amount of distortion in the overlapping areas (edge of the flight path), they are not well suited for trying to merge them later on.
GSD should be around 2.5 cm according to DJI flight planner

I guess I misunderstood the fast-ortho option. I thought all it did was skip dense point cloud reconstruction and DEM . Which would as far as I understand result in somehwhat less precision when de-distorting the images. I just saw, that selecting the Fast Orthphoto preset defaults the GSD to 5cm. Do I understand it correctly if I set it to 1 but my recorded resolution was 2.5, the resolution will be 2.5 cm/px at the end?

Thanks for your help

Yes, it will give you the flown GSD of 2.5. I don't have any experience with the following plugin, but maybe you can try. First you process each subset and then merge them with : https://github.com/uav4geo/QRasterMerge

Only compatible with qgis 3.x

With split set to set to 1000 images it finally worked. But it also took 35 hours.
I also had to realize that the whole orthophoto gets too big for further processing. There are also areas that didn't get reconstructed well (densly forested without enough matches)
I am now testing different ideas for the individual flights.

Thank you guys for your help