Topics

1

General / GPU does not work at build-dense-cloud in network cluster processing

« on: April 20, 2018, 06:28:38 AM »

Hi,all
I am trying network cluster processing. I gathered all machines in the same  room. The network load between node and storage is  880 Mbps. I am doing batch processing.

But the GPUs does not  work at build-dence-cloud.  They work at other phases.

What is wrong?

2

General / Remote cluster processing is much slower than stand alone

« on: April 18, 2018, 08:51:18 AM »

Hi all,
I tried a remote cluster processing. Site 1 is consisted with a server, a client and a node in one computer. Site 2 is a node. Site 3 is a file storage. These sites are connected with VPN and optical fiber, and several kilometers off each other.
In 200 photographs case, it takes 10 hours to get orthomosaic, but it takes only 1.5 hours by site 2 stand alone computer. Photo matching is very time-consuming in the remote cluster processing.

What can I do to process faster than stand alone?

OR

Do I have to get all computers together in the same building to process faster?

3

Python and Java API / How to get values of estimateImageQuality() in version 1.4

« on: April 18, 2018, 07:48:50 AM »

Hi,
When I select photos before alignment, I used a code like below.

Code: [Select]

import PhotoScan

for chunk in PhotoScan.app.document.chunks:
    chunk.estimateImageQuality(chunk.cameras)
    for camera in chunk.cameras:
        if float(camera.photo.meta["Image/Quality"]) < 0.7:
            camera.enabled = False
    chunk.matchPhotos(accuracy = PhotoScan.MediumAccuracy)
    chunk.alignCameras()

But after I updated PhotoScan version from 1.3 to 1.4,  this code does not work.
camera.photo.meta[Image/Quality] does not exists after estimateImageQuality().

How can I get/compare "Image/Quality" value?

Is this my misunderstanding??