Messages

31

General / Does increasing the exposure value of the dark images help?

« on: July 17, 2018, 01:07:41 AM »

Hi,
I have two questions. The first question is about whether Photoscan adjusts or stretches the histogram of the images for the reconstruction. If it does not, does reasonably increasing the exposure value of the low-light images help reconstruct a better 3D model? Does applying curves to the image histogram by increasing the shadow brightness and decreasing the highlight brightness, thereby creating an HDR image help reconstruct a better 3D model?
Second, building HDR images using auto-exposure bracketing shots and thereby increasing the dynamic range help reconstruct a better 3D model?

Thank you!

32

General / Re: How can I manually align images using features in the images, like Reality Captu

« on: November 16, 2017, 05:23:35 AM »

This video helps very much as well. She explains so nicely how to use markers to align multiple chunks.

https://www.youtube.com/watch?v=pYE0KOa5oGk

33

General / Re: Moving points in a chunk moves the points in other chunks correspondingly

« on: November 15, 2017, 08:13:42 PM »

I am sorry but it's weird. It happens again. I simply can't figure out how it works.

34

General / How to remove the points some way off the dense reconstruction surface?

« on: November 15, 2017, 08:05:07 PM »

Because we are taking pictures of the surface, the reconstructed points should have a sharp and dense surface. Then the points some way off that dense surface should be regarded as a reconstruction error or the noise introduced by sky or something else, if we don't have much relief on the surface such as high relief rocks or the vegetation. Do we have a tool to remove those points that are 3-sigma off the dense surface?

35

General / Re: Moving points in a chunk moves the points in other chunks correspondingly

« on: November 15, 2017, 07:58:32 PM »

I just wanted to update it.
I aligned the chunks at the same time using "Workflow-Align Chunks" and choosing all the chunks together. Then the points in each chunk appear near the center though moving the points in one chunk still moves them in the other chunks.

36

General / Moving points in a chunk moves the points in other chunks correspondingly

« on: November 15, 2017, 03:45:01 AM »

I have 5 chunks in my 3D scan and each chunk has been taken at the different distances from the object. The photos in the first chunk were taken very far from the object and cover a larger area around the object. The photos in the second chunk were taken a little bit closer to the object and cover a smaller area around the object. Likewise, each of the five chunks contains photos taken from the different distances from the object and cover the different areas around it.

I have aligned each chunk separately and found that every time I move the tie points in a chunk, it moves the tie points in the other chunks accordingly. If I place the points in the first chunk to the center of the view where the trackball is, then the points in the other chunks are moved off the center of the view. So, I have not been able to place the tie points in all chunks at the center of each chunk's view at the same time.

Why does moving tie points in a chunk affect the location of the other chunks' tie points? How can I link or unlink between them? How can I place them at the center of each chunk's view?

37

General / Re: How can I manually align images using features in the images, like Reality Captu

« on: November 13, 2017, 12:14:09 PM »

Thank you so much! It helped very much!

38

General / How can I manually align images using features in the images, like Reality Captu

« on: November 13, 2017, 01:48:31 AM »

I am relatively new to Photoscan, so I am afraid that it could already be included in the Photoscan functions. But I was not able to find this function in Photoscan yet.

I am not talking about the alignment using markers or geotagged images. I don't have any GPS data for the images. I am talking about aligning images by using the common features in the images.

Today I used Reality Capture trial to reconstruct the models using images taken by UAV. For the first tryout, the reconstruction was not successful. Then I tried to manually align the images by adding control points on the common features of each image. I picked a couple of images from each chunk and visually marked the common features found in each of the images. Then it gave me a much-improved result in a very short period of time.

Please don't give me an answer regarding "Using GPS based ground control points". It is useless if I don't have GPS data or if an object is too small, in which case GPS data have no meaning. Thank you!

* I've attached the screen capture for aligning images using control points.

* The Photoscan version I am using is 1.3.4.

39

General / Re: Only 10 out of 1280 CUDA Cores are being used by Photoscan

« on: October 18, 2017, 06:41:53 AM »

Thank you Smallpoly for the attachment.
I am sorry guys, I am late to follow it up.
 
I got this message from the Agisoft customer center.

"Have you checked the utilization of GPU using NVIDIA SMI? Or you just looking at the number of processor units in the GPU.

According to the GPU specification (http://www.guru3d.com/articles-pages/msi-geforce-gtx-1060-6gt-oc-review,5.html) The GeForce GTX 1060 has 10 x 128 shader processors makes a total of 1280 shader processors, so number 10 here doesn't mean that PhotoScan is using only 10 CUDA cores."

The website above describes that
The GeForce GTX 1060 (GP104-400) has 10 x 128 shader processors makes a total of 1,280 shader processors.
The GeForce GTX 1070 (GP104-200) has 15 x 128 shader processors makes a total of 1,920 shader processors.
The GeForce GTX 1080 (GP104-400) has 20 x 128 shader processors makes a total of 2,560 shader processors.

So they say that "using 10 cores" means that the software is using 10*128 = 1280 shader processors. Does it make sense?

40

General / Re: Is the PhotoScan stuck?

« on: October 17, 2017, 06:12:17 PM »

I aligned and merged two chunks and then started building a dense cloud with "HIGH". Two chunks contain about 250 pictures from Phantom 4.

41

General / Is the PhotoScan stuck?

« on: October 17, 2017, 05:58:39 PM »

I just wanted to know how I can decide if the Photoscan is stuck or not by just looking at how my PC is running right now. The title bar of the progress dialog box of Photoscan shows "Not Responding" after I clicked on it and it does not show any changes in numbers and progress bar. But the task manager shows it is using about 86% of CPU and 6GB of RAM out of 32 GB, and the statistics of the CPU usage and the RAM taken is changing a little bit (85% +- 7%, 6.372.7 GB +- 1GB). I can also manipulate the other window stuff including small applications like notepad. When the Windows looks running and the Photoscan is still changing its usage of CPU and RAM the progress dialog box looks frozen. In this case, how can I know if it is stuck or not? Do you think it is stuck?

42

General / Photometric correction

« on: October 15, 2017, 02:03:14 AM »

Alexey said before that small changes of contrast, brightness and etc. wouldn't affect the geometry reconstruction process. Then, what about the having bracketing exposures and making hdr images? I think it will reduce the noise from the darker portions and will help the reconstruction. How much do you think it will affect the quality of the reconstruction? Thanks,

43

General / Re: Storing color by vertices or in a texture

« on: October 15, 2017, 02:00:45 AM »

Now I have a better understanding of 3D modelling terms and workflows and it solved my question. Your answer is much appreciated; it helped me understand it better. Thanks!

44

General / Only 10 out of 1280 CUDA Cores are being used by Photoscan

« on: October 15, 2017, 01:55:42 AM »

I am using Photoscan Pro 1.3.x. In Preferences - GPUs, I found that Photoscan set only 10 cuda cores to be used and I can't change it, although my graphics card GTX 1060 6GB has 1280 cores. How can I increase the number of the cores that will be used? thanks!

45

General / Storing color by vertices or in a texture

« on: October 09, 2017, 06:43:47 AM »

Hello,
I am relatively new to this software and the field of  photogrammetry. I heard that the color information can be sotred either by vertex (colorize) or in a texture, which I don't understand very well. Could you please let me know how I can choose one of two ways of storing color, and in what steps of the workflow in Photoscan I can do that? The document I referred was about the reconstruction using Reality Capture, and I would like to do the same workflow using Photoscan! Thank you!