Messages

46

General / Re: Hardware/Processing optimization

« on: July 28, 2014, 10:54:03 PM »

Align Photos does not use GPU.
Create Dense Cloud uses both CPU and GPU.
Build Mesh does not use GPU.

"Disabling" CPU cores does not actually disable specific cores. All cores continue to be used. I experimented with "disabling" various numbers of cores and did not find major differences in execution time. The AgiSoft advice of "disabling" 1 physical core per GPU appeared to be best. For a i7 4 core / 8 hyperthread CPU that means disabling 2 of 8 "cores".

47

General / Re: Hardware/Processing optimization

« on: July 27, 2014, 01:24:42 AM »

My understanding is PhotoScan observes the actual performance of the CPU and GPU, and then adjusts the load on each to optimize overall speed.

48

General / Re: Macro Photoscan

« on: July 25, 2014, 07:12:41 PM »

Very nice.

What camera and lens?  Did you use focus stacking?

49

General / Re: Having difficulty with model of amphitheater; please help.

« on: July 25, 2014, 07:09:52 PM »

Edit:  Posted to wrong thread.

50

General / Re: What's wrong in this model?

« on: July 24, 2014, 11:32:57 PM »

What settings are you using for Build Mesh?
- Surface type?
- Source data?
- Interpolation? (under Advanced)

Try:
Surface type:  Arbitrary
Source data:  Dense cloud
Interpolation:  Enabled

51

Bug Reports / Dense Cloud Build - exponential time increase

« on: July 23, 2014, 06:08:14 AM »

Times for "filtering depth maps" during Build Dense Cloud appear to increase exponentially with the number of cameras used on my current project. Is this normal and expected? I don't recall encountering it before.  Attached is log file for a batch run with three chunks using 361, 402 and 502 cameras. (File edited to reduce size so it could be uploaded) Times for "filtering depth maps" were 626 sec, 2332 sec and 4484 sec.

PhotoScan Standard Edition Version 1.0.4 build 1847 (64 bit)
Windows 8.1
i7-3770 CPU, 4 core and 8 virtual threads, virtually 100% use of CPU during "filtering depth maps"
AMD 7770 GPU, not used during "filtering depth maps"
16GB memory, 8.4GB maximum used during run.

GPU is used during "depth reconstruction".

52

General / Re: single line processing

« on: July 23, 2014, 05:38:31 AM »

And there is sufficient "texture" in each photo.

53

Feature Requests / Re: Manual photo alignment in Standard edition

« on: July 21, 2014, 07:00:02 AM »

Photos for PhotoScan have fundamentally different requirements than photos for panoramic stitching software such as PTgui.   Photos whose overlap and visible features are more than adequate for panorama stitching may not be suitable for PhotoScan

1. Photos for PhotoScan need to be taken from different positions. PhotoScan won't work if the photos are taken from the same location.

Panoramic stitching needs photos taken from the same location.

2. Parts of an object which are to be modeled in PhotoScan needs to appear in a minimum of two photos taken from different positions, and photos taken from three or more positions are much better. This means for a single row of photos an overlap of at least 50% is needed, and overlap of ore than 67% is preferred.

Panoramic stitch only needs enough overlap to align the photos relative to each other. 20% is typically desirable but I've stitched photos with less overlap into panoramas.

3. In general PhotoScan needs photos with more "texture" than panoramic stitching software.

54

Feature Requests / Re: Manual photo alignment in Standard edition

« on: July 21, 2014, 06:41:11 AM »

I've found "Reset Camera Alignment" and "Align Selected Photos" to be very useful for photos which do not initially align or align incorrectly. After initial alignment right click on a photo thumbnail to open the menu with those commands. Frequently but not always a photo will align correctly when these are used. Also, I've had sets of photos which did not completely align with Accuracy set to Low or Medium align with Accuracy set to Medium or High.

Also, occasionally photos which won't align with Pair Preselection enabled will align with it disabled.

55

General / Re: When to merge and align chunks?

« on: July 21, 2014, 12:00:25 AM »

Similar to this question, what exactly does align chunks do?  I tried a test and had very confusing results. 

I took images of the same place with two cameras.  I put the images from both cameras into separate chunks.  I ran the workflow on each chunk, then aligned the chunks and exported orthophotos from each chunk.  The orthophotos don't overlay very closely when loaded into an image editor.

Next, I loaded all images into a single chunk and aligned the photos.  Then I exported two orthophotos for each group of images.  The output orthophotos exactly overlay.

So exactly what is the intended benefit of align chunks if the resulting images don't actually overlay properly?   The ideal scenario for me is that I load images from different cameras into different chunks, align chunks, and then the exported orthophotos perfectly overlay so that I can do image processing on them.

How many photos? How much overlap?

What settings were used for "Align Photos"? What is the "Effective Overlap" and "Reprojection Error" for "Show Info" for each chunk.

What settings used for "Align Chunks"?

56

General / Re: Beginner with some basic questions

« on: July 18, 2014, 05:40:01 AM »

The target objects must have some random visual texture. Bones should be fine. Clean plastic and paint generally won't work. I'll be very surprised if you get anything with Solo cup. Find a rock or tree branch and use that for your experiments.

Masking is needed if the background is changing or moving. Otherwise masking is not needed. If the background has a random visual texture and is not moving or changing then not masking may result in better alignment.

Don't worry about markers yet. Learn how to take and align photos without adding complications.

There is no requirement to take photos all the way around an object. PhotoScan works with photos taken of only one side of an object. But of course you only get results for the side of the object visible in the photos.

Find a rock, tree branch or other object with a random visual texture (emphasis on random). Practice with it until you're getting good results.

57

General / Re: RAW, JPG and different types of noise

« on: July 16, 2014, 11:13:38 PM »

Frank, try developing the raw files as JPEGS with 100% quality.

I used to always use 16 bit TIFF images as input to PhotoScan. The images were obtained by "developing" RAW files. But after some experimentation and comparisons I switched to using JPEG images with "quality" of 100%. These images are also obtained by "developing" RAW files, and are not the JPEG images output by the camera. My comparisons showed that using 100% quality JPEG images results in a decrease of a few percent in the number of sparse points, and does not have a noticeable impact on the PhotoScan results. JPEG images with lower quality settings are smaller but result in fewer sparse points and sufficiently low settings of quality resulted in noticeable impact on PhotoScan results.

All JPEG images are compressed, and as RalfH noted above the compression can result in artifacts and loss of data. But not all JPEG images are the same. The algorithm used for generating JPEGs has an input parameter called "quality" which can range up to 100%. Higher quality values result in fewer artifacts, less loss of detail but larger files. Generally the quality setting used internally in a camera to generate JPEGs is not available but is selected to keep file sizes relatively small.

16 bit TIFFs from raw files from my Canon T1i/500D are 86.3 MB. JPEGs with quality set to 100% range from 10 MB to 21 MB depending on image quality. (The corresponding JPEGs direct from the camera with unknown "quality" setting range from 4 MB to 6 MB.)

Creating the JPEGs by "developing" the raw files allows me to correct for chromatic aberration, adjust the exposure and contrast if needed while avoiding clipping in the areas of interest, and select the JPEG quality setting.

58

Feature Requests / Re: Copy Camera command

« on: July 15, 2014, 07:18:58 PM »

My current work-around is to duplicate the chunk I need to copy photos from, move the photos from the duplicate chunk, and then delete the duplicate chunk.

59

Face and Body Scanning / Re: Picture Style Setting

« on: July 14, 2014, 08:36:41 PM »

My guess is the sharpness adjustment in the camera is applied after before the image is compressed and converted to JPEG. Otherwise the individual pixel values would have to be extracted from the compressed JPEG data, sharpness applied, and then the image recompressed into JPEG format. It would be much more efficient and quicker to apply sharpening before compression.

Have you considered capturing the images in RAW (.cr2) as well as JPEG (.jpg) format? You could then process the images in the DPP software which came with the camera. DPP includes the ability to change or modify the "Picture Style".

60

Face and Body Scanning / Re: Picture Style Setting

« on: July 14, 2014, 02:50:51 PM »

Sharpening an image can cause loss of detail. Sharpening adjusts the local contrast to make the images appear "sharper". It does not increase the amount of detail in an image.