Messages

1

General / Re: Slow build mesh step

« on: October 30, 2015, 01:09:53 PM »

I have checked and I have  16GB of memory of which 11 GB of it is unused. So I'm inclined to say I'm not running out of memory.

2

General / Slow build mesh step

« on: October 29, 2015, 04:56:01 PM »

Hi,

I've noticed that the PhotoScan build mesh step sometimes gets incredibly slow (from ~1 minute to dozens of minutes) every now and then. It also seems like it behaves on different computers differently for the exact same set of images and data. Does anyone know what can possibly be the cause for it? It's quite annoying since you never know when it'll slow down again....

thanks.

3

Python and Java API / Mixing cameras with and without calibration

« on: July 20, 2015, 05:24:12 PM »

Using the PhotoScan targets, I have calibrated a set of 9 cameras to scan an object.

This calibration allows me to have the correct scale for the scanned objects.

I have another set of  9 images, taking by the same cameras but with a different object orientation (the background is featureless so that's not a worry)

I now want to use all 18 images to build a more complete object model that is scaled correctly. I have tried importing all 18 images and then importing the camera calibration file (containing 9 camera information) but this only seemed to use 9 cameras still.

Any idea how to do this in PhotoScan?

4

Python and Java API / Importing Cameras and having correct scale

« on: July 16, 2015, 04:48:26 PM »

Hi,

I have a fixed set of cameras that I calibrated using PhotoScans target calibration procedure. I exported the cameras as an XML and file and verified all cameras are in the XML file.

Next, I have another set of images taken with the same setup, and now I just want to import camera locations, skip the alignment step (since I already know how the cameras are precisely located) and have the resulting end model with the correct scale.

However, after running

Code: [Select]

chunk.importCameras("/some/path")

I still need to run alignCameras. Furthermore, the resulting OBJ model after running dense cloud and mesh creation, is not scaled correctly.

If however I import the cameras using the GUI, then it automatically prompts me a dialog to align the cameras but the resulting .obj file has the correct scale this time!

My questions are:
- How can I skip the alignment step and directly generate sparse point cloud since I have imported camera locations?
- Why is there are difference between GUI import cameras and importing from Python? How can I make the Python side work correctly?