WebODM Progress Indicators & Scaling for 3k+ Mavic 3 Images

yeah for that many photos, that is not enough ram...

I've got another stick of 32 GB coming in, but that's as much as I can afford right now given the RAM prices. I am using split = 200 and overlap = 10. Our imagery overlap is 90%. I set the octree depth to 11 and mesh size to 600000 per the ODM best practices.

you could reduce the settings and had harsher point removal to make sure you don't run out of ram. otherwise might not work. I think last time I ran that many I was over 120gb of ram with high settings.

Is there anything like "The Idiots Guide to WebODM"?

I’ll hop in to the topic… total beginner in webodm here - how do you think, how much ram should i have to process around 6500 photos? 🫣

@szymoon @JWX Here is the hardware requirements from the ODX docs

Is ther ea link to that book? I tried to get Claude Code to help me explain stuff... it was wildly wrong so I think I'm done with that guy lol. That was so frustrating.

https://discordapp.com/channels/1490218202033029212/1490218203807355004/1514428363878694983

Thank you!

@JWX this is meant for webODM lighting, but it still helps explain things for most of the task options for self hosted: https://docs.webodm.net/references/task-options/

That's really useful too. I like that there's a simple calibration tool too.

Well, the 5070 is really paying off with the depth maps. Those were massive and slow before.

Is there a way to get WebODM to have more of a GUI to show cameras, points, footprints, etc.? I'm thinking that would be helpful for figuring out why I saw some odd smearing in one part of the model. I'm afraid I'll have to send out the drone guy again.

what are you running to render? are you just running the processing node or the whole stack for viewing? can you send a screenshot please?

Yeah! Here is what I have. I'm still pretty new to Docker and nodes. I've been learning a little bit of that through this project which is kinda helping me at my day job.

Also, that is deployed locally.

you can view camera positions in once the render is complete in the 2D and 3D views.

Top right for maps, top left for models

Thanks! @LIL-Diabeetus

No problem!

I'm getting some horrible solutions.

My parameters were:

• auto-boundary:true
• sfm-algorithm:triangulation
• sfm-no-partial:true
• use-hybrid-bundle-adjustment:true
• pc-quality:high
• skip-3dmodel:true
• mesh-octree-depth:11
• mesh-size:600000
• rerun-from:dataset
• split:200
• split-overlap:10

@JWX I am out of the office right now, but I can give you some presets that I think would work good for you. If you give me a few hours I should be back. First things to try is changing:

• pc-quality to medium (less dense pointcloud, better for lower ram)
• pc-filter: 2
• orthophoto-resolution:2

everything else seems to look good though for settings at a glance

@LIL-Diabeetus I can wait awhile. Thanks for getting back to me. I'm just not sure why I'm getting those blury parts when it seems like there are plenty of images over that area.

Also make the sfm-algorithm to incremental instead of triangulation

@LIL-Diabeetus, I think I found part of the problem. A bunch of photos were recorded at -45 and -135. I dropped anything not -90 and that cut the dataset to 582 images. I'm using most of your recommended settings. I missed the pc-filter:2.

Sounds good! let me know how it turns out!

@LIL-Diabeetus dude! It's awesome! 1.25 hours, 2 cm resolution.