Textured model optimization


#1

Does anyone have any ideas on how to get the 3D portion to display faster, I have a project with 150 images, and the 2d looks great, and the 3d point cloud is barely ok, but if I select the Textured model button it takes forever, like an hour or so, then sometimes it crashes the viewing computer.

I have it loaded on Ubuntu in AWS and with these settings: 4.2xlarge type and has 61GB RAM and 8 vCPU

Has anyone had experience optimizing this?


#2

Part of the problem could be due to the size of the texture images (check the size of the png files in odm_texturing, you’ll see why). When accessing the models over the internet you probably want to do a combination of texture resizing and/or decimation (to reduce the size of the OBJ), or even better generate different LODs and stream those incrementally. This is how sketchfab does it.

For your purposes, perhaps simply resizing the size of the texture files might be sufficient.


#3

We have an item related to this open in https://github.com/OpenDroneMap/WebODM/issues/383, but haven’t had time to implement it yet.


#4

How would I go about doing those things, it there some documentation on all 3 of those things you mentioned. I hate to bug you for each item on how to set up, and how to configure.

Honestly, is there a way to export and then load into sketchFab does it?

Last week I tried to export the obj into meshlab and it crashed 2 different systems.