Topics

1

General / New Depth Map each mesh generation

« on: August 24, 2023, 06:46:20 PM »

Even though I have "keep depth maps" checked off in preferences, Metashape generates a new one each time I generate a new mesh (same point cloud different resolutions). Depth Map generation is easily half the total processing time. Is this the way it's suppose to work? I would think that once I've generated the dense point cloud any mesh is going to be a derivative of that and a new depth map shouldn't be required. If I use the point cloud to generate the mesh I get a warning that it wont be as precise....

2

Python and Java API / Gradual Selection Python Script

« on: April 01, 2020, 08:12:44 PM »

Hi All,
I really need a python script to go through the gradual selection process in between camera alignment and dense cloud. There’s nothing in the list of programs on the wiki or github and although I’ve seen some discussions in the forum it’s fragmented and not a complete procedure. While many years ago I use to program in Pascal and Basic, I do not know python well enough to do this on my own. Thanks for any help I can get.

3

Python and Java API / Split into Chunks Dense Cloud w High Quality Setting w Aggressive Filtering

« on: July 28, 2019, 08:24:26 PM »

I've looked over the Split_into_chunks code and I'm not sure where and what I have to change to generate a high quality dense cloud w aggressive filtering.

Under the section - "if buildDense :" there are several if- then statements and I don't program in python...yet.
Isn't there somewhere at the top we could declare the constants and leave it at that?

Thanks for your help

Edit:ok, there's an unmarked button to define the quality...I missed that. Sry

4

Python and Java API / Orientation in Split into Chunks

« on: December 31, 2018, 10:35:22 PM »

I have a model that is too big for my computer so I am going to use the Split_into_Chunks python script to process it. As currently oriented, I want to break the model into three chunks along the z axis (1X3). From reading over the code I get the impression that the program generates chunks along the X and Y axis. Reorienting the model is not a big deal but I thought I would confirm this before making any changes. If it is x and y, is there any easy way to mod the code to generate along the z axis?

Also wanted to check if, after updating to MetaShape if I need to change the references to Photoscan in the code to Metashape ex. "import Photoscan" to "import Metashape". I imagine so.

5

General / Accuracy and precision in Photoscan

« on: August 16, 2016, 05:08:44 PM »

I’m trying to determine a metric for the accuracy and reliability of measurements taken off of a photogrammetric model created in Photoscan.

If I take a picture using a 24mp camera and a 60mm lens from 1 meter away the best resolution I will have is approximately 0.2mm. That’s the size (0.2mm^2) of a pixel normal to the camera and I assume the smallest piece of information I can obtain.

When I take 500 of these pictures and compile them into a photogrammetric model using photoscan, I get a point cloud with point spacings of less than 0.05mm. There is clearly a fair amount of interpolation and extrapolation going on.

The pixel size times the rms reprojection error I think would fairly represent the most accuracy one could expect but I have seen details in the models that should not have been resolved if this were the case.

I assume, the initial point cloud in photo alignment effectively establishes benchmark locations over the model then the dense cloud creation carries that process further based on the imagery information but at what point does extrapolation and interpolation take over from accurate determination of xyz coordinates based on the photo comparisons and how accurate are those point determinations.

Any thoughts on this would be appreciated. I get the feeling at times that the precision of the models far exceed the accuracy of the base information. Like using a tape measure and recording the results to the 4th decimal place.

6

General / Adding independent graphics card for performance boost

« on: July 02, 2016, 06:38:09 PM »

I'm currently running an AMD R9 290 graphics card in my PS computer. I have an older graphics card AMD 6950(?) doing nothing. If I add that into the PS computer as an independent graphics card (obviously not crossfired) will I get a performance boost or will I just screw things up? I have both the power and cooling capacity to support it.

7

General / Upgrading for speed

« on: June 17, 2016, 06:45:53 PM »

It’s clear from my growing experience that I need to upgrade my system to get processing results in a reasonable amount of time. I’m currently using a windows system with an i5-4690k @ 3.5ghz with 16gb of memory and an AMD R9 290 graphics card. Increasing memory to 32gb is a must, but which would give me a better performance boost, adding another graphics card or replacing the CPU with say an i7-4790 4ghz.

8

General / Manual Photo Alignment and adding photos

« on: June 17, 2016, 06:28:14 PM »

I read in an earlier post somewhere that excluding detail shots from the initial alignment and development of the sparse point cloud would speed up the process and sometimes unblock a choke point in the processing and that once the sparse cloud had been developed the detail shots could be added in using manual alignment. I tried this but the manual alignment (align selected camera) wouldn't work. Reading further I found a post by Alexey that explained that manual alignment looks for previous match points so therefor only works if the manually aligned images were part of the original set.

Is there a way to add and align images to the original sparse point cloud?

If the images are added after the sparse point cloud is created and I go straight to dense point cloud development will the new images be incorporated?

9

General / Two noob questions

« on: June 07, 2016, 09:07:45 PM »

Just starting out with detailed modeling:

1. When I load and align my images they are inevitably grossly skewed/rotated from the trackball/local coordinate system. Its easy enough to translate the model to center on the trackball but I haven't figured how to align the trackball axis with the preferred model axis. What am I missing?

2. For photo alignment, I want to use all the capture points in the images. Once a preliminary model is created I want to use it to create a mask and use that mask to regenerate (hopefully) a better dense point cloud. There is a button for using or not using masks on the photo alignment but there is no equivalent switch on the dense cloud generation. Does the program assume that if masks exist then it will use them?

Thanks

10

General / Nikon D600 or D700

« on: May 17, 2016, 09:46:12 PM »

Is it better to have more pixels or higher quality pixels? I have a project to create a model of a statue that must be shot in low light conditions. I have a choice of a Nikon D600 at 24 megapixels or a Nikon D700 at 12 megapixels. Both have the same size Nikon FX CMOS so one would expect less noise on the D700 at the same aperture and shutter speeds. I did a test earlier (on a different camera) and was surprised at the impact that noise from different ISO settings has on the final model just going from ISO 100 to ISO 200. Of course there, the pixels were all the same size whereas here I have a choice of more pixels or less. I'm hoping that someone more familiar with how the pixel data is interpreted can give me some insight.

11

General / Masks and Dense Cloud Creation

« on: April 10, 2016, 12:48:11 AM »

Hi,
New to Photoscan (on trial), so far getting great results. My question is, do masks used for the dense cloud have to be prepared for all photos or can a representative photo that contains the area to be masked be used, and it will be carried through out the model? I don't particularly want to prepare a mask for each photo when there are hundreds involved.
Thanks in advance