Large Point Cloud meshing solutions


(Michal Zurawski) #1

Hi guys,

It has been a while I’m thinking what to do when I will face huge point cloud that I’m unable to mesh at once. I use Agisoft Metashape and i simply run out of RAM with my latest project. I’m looking for some ideas or external software that allows me to create decent mesh.

The problem is, that this 300M point cloud is not enough detailed.

The amount of details is nice but still not high enough to fully analyze surface of the object. The next approach I use 135 prime lens, that will produce sharper pictures. I did this scan with 70mm, so using 135 will double at least amount of photo used so point cloud will be even more dense. That is why I need solution. Your ideas are kindly welcome.

(Cie) #2

Hi Michal,

Not sure if this helps, but have you tried the new ‘generate mesh from Depth maps option’ within Metashape? You don’t need to generate a point cloud first this way. This process uses less RAM, so you might be able to get it this way. Only thing is i’m not sure the detail will be good enough, i’ve seen fluctuating results with this new option.


(Michal Zurawski) #3

Same. Basing on my experience It saves about 20% RAM usage. I’v tried already with no luck. I was able to generate successfully model in high-field mode ( but this is not the issue. One day (soon, I have already data set) I will have need to use arbitrary model and then project is dead. I don’t know if it is true but I’ve heard that Reality Capture can do meshing better.

Thanks anyway mate

(Nebulousflynn) #4

Have you tried the RealityCapture demo? This will quickly tell you if it provides what you need :slight_smile:

I use both Metashape and RC depending on my project. It is also possible to export cameras and images from Metashape to open in RealityCapture for mesh processing.

(Michal Zurawski) #5

I will try to use the demo. The problem is that I registered for a demo and… I havent istalled the software… and one month has passed -_-. Anyway, i’m trying to prepare my model as you did. This is an architecture interior so i’m gonna try to publish point cloud. We’ll see