Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Andrew

Pages: 1 2 3 [4] 5 6
General / Re: AMD RADEON slow response big dataset
« on: December 21, 2013, 05:02:21 PM »
Thank you that Alexey, missed that thread :) Will see if that helps.

General / Re: AMD RADEON slow response big dataset
« on: December 21, 2013, 04:50:56 PM »
Good catch, I never gave it a thought but now I took a closer look and my 7970s also report 2GB in agisoft, despite being 3GB cards.  :o

I remember having similar problem with Autodesk Mudbox, where they suggest manually changing some (Mudbox-specific) System Variable in Windows...

I sure hope there is a way to reclaim that 1/3 unused memory in Agisoft...

Feature Requests / Adding new photos to existing chunk
« on: December 01, 2013, 03:05:36 PM »
Adding new photos and aligning them to already existing aligned chunk does not seem to work, could that be made possible?

Every now and then, after grabbing photos and aligning them in Photoscan I realize I didn't get enough coverage of some area, or I messed up cam settings and some photos are unusable. Ideally I would like to take additional photos and just add them to existing chunk, just align those new photos to the previously aligned ones. This could easily save hours, as usually hundreds of photos are just fine and I only need to add a couple new ones.


General / Re: Calculating memory usage
« on: November 10, 2013, 10:53:59 PM »
Isn't that page out of date though? I believe that PS 1.0.0 brought significant changes (dense point cloud) and some reduction in memory requirements? I never ran any comparisons between old PS and 1.0.0 prerelease, but it certainly felt I could deal with larger projects than previously.

76 post views and no answer, unless people in the know missed this thread, I guess it's safe to assume that Photoscan Professional Edition does not provide mesh quality edge over Standard Edition in non-DEM applications...

I have quite a lot experience with PS Standard Edition (and I'm really quite happy with results!), but I keep wondering - can Professional Edition enhance reconstruction quality even further?

I know it can help streamline process with ability to scale/orient meshes properly and with broad Python scripting capabilities, but I am curious about features that refine/optimize reconstruction - do these only make sense for DEM applications or will they also result in more precise/smoother reconstructions of human faces and such?

Perhaps anyone can share screengrabs from same dataset processed with and without ProEdition-specific features?


General / Re: Photoscan texture generation vs DOF
« on: September 09, 2013, 12:03:08 PM »
Seems this thread got burried before anyone from Agisoft (or anyone else with knowledge) could chime in.

To recap my main question: is there a way in PS to detect which photo was used to project specific part of textured scan?

The reason for this is, due to difficult shooting conditions, my photos occasionaly have too shallow DOF to capture everything in focus. I need whatever is sharp in these photos for good reconstruction, but I often need to discard out of focus areas by tedious masking. Not every OOF image part needs to be masked out, sometimes PS makes a good call and picks from sharp photos, so masking out every OOF pixel in all images seems like a waste of time - If only I knew which photos PS chooses for texturing specific areas.


General / Re: Photoscan texture generation vs DOF
« on: September 04, 2013, 05:06:55 PM »

In real world scenarios there are situations where photos contain both vital sharp areas, and areas where parts of your subject are out of focus. High f-stop and smart shooting plan can minimize this, but often can't elliminate completely.

So disabling entire photo during texture generation is not an option.

I know that with enough manual masking, areas out of focus can be excluded from texturing, but most of them do not actually affect texturing, only some do, so perhaps there are ways to point which photo contributed blur in the texture, or have PS help disregard blurry photo areas (blur-aware texturing algorithm, masking tool etc.)


General / Photoscan texture generation vs DOF
« on: September 04, 2013, 01:30:31 PM »
When acquiring complex sets of shapes, I often end up grabbing some parts in perfect focus while other parts are blurred by DOF (too close or too far from focal plane). However, I always make sure that every square inch of my project is perfectly sharp on at least a couple of photos.

Problem is, Photoscan doesn't always use these sharp photos for texture generation. I try to mask out DOF areas by hand, but on larger datasets this becomes tedious.


1. Is there any way of knowing which photo photoscan used to texture selected area?
2. PS ver 1.0.0 has this 'Estimate image quality' feature to quantify image sharpness (or perhaps it quantifies other qualities of a photo?), does this mean PS can/could use this data to pick sharpest photo parts for texturing?
3. Could it be possible to use image quality data to automatically (with adjustable tolerance) mask out blurred areas in photos?


General / Re: fixed focal length ?
« on: September 03, 2013, 11:07:35 PM »
Logic suggests its best to stick to fixed lens or min/max zoom, though I wonder: does splitting calibration settings for photos shot at different focal lengths result in noticeably worse reconstruction?

Real world example: when capturing a monument, I can shoot it from most directions from the same distance @18mm, but there's an obstacle from one side, preventing me from shooting up close - am I better off shooting still @18mm with monument being rather small in frame, or can I zoom in and not worry about different calibration params for that photo?


Andrew, I was just wondering regarding this point about the 600D:
- syncs slower than Nikon, occasional lag inconsistencies

I thought from your testing with the Camera Axe/Multi Flash Plus that the 550D was inconsistent but the 600D was not (when pre-focus command was taken into account)?

My tests were not very accurate (I measured against incredibly slow cheap studio flash), so I could only confirm that they are much more reliable than 550D, but I would have to get much faster speedlight and do extensive tests to get more confidence in my findings. In contrast, Lee has been testing and using 600d's in real work scenarios for a good while now and I have a lot of confidence in his claims, so I based what I wrote on his findings (big thanks, Lee!).

Oh, I haven't noticed particularly slow Smart Shooter transfers from D3200, no. Photos take a tiny bit longer (perhaps 20-30% longer) to land on my HDD but files are larger so thats understandable.


My summary of this thread (feel free to add/correct if I missed something or if I got anything


Canon 600D: (from my own experience)

+ excellent image quality
+ reasonable price
+ standard USB connection with decent cable lenght
+ wide range of lenses, including good but affordable ones

- syncs slower than Nikon, occasional lag inconsistencies
- can't fully open battery bay while mounted

Nikon 7100D: (from my own experience)

+ excellent image quality (virtually same res as 600D, slightly better dynamic range)
+ battery bay fully accessible when mounted
+ fast shutter, syncs better, more consistent

- 2,5 times the price of 600D
- larger filesize=Photoscan needs more memory and time to produce results identical to 600D
- silly proprietary USB cable (at least decent lenght, compared to D3200)
- occasional build quality issues (faulty USB port, oil spills on sensor, defective focus)
- somewhat more expensive lenses

Nikon D3200 (from my own experience)

+ significantly cheaper than 600D
+ syncs faster than 600D
+battery bay fully accessible when mounted

- produces very soft images unsuitable for serious photogrammetry
- silly proprietary USB cable, absurdly short

Nikon D800/E

+ super speedy and syncs fast
+ good image quality (although supposedly does not edge 600D despite 36mp vs 18mp)
+ excellent build quality

- extremely expensive
- full frame means DOF issues
- full frame means expensive and less flexible lenses
- oil spills soil the sensor

Here goes another crop comparing D3200 vs 600D vs D7100, full images available upon request:

Conclusion is obvious: Considering price, quality and performance, Canon 600D remains as the

most reasonable and most widely recommended DSLR for multicam photogrammetry. A camera that is 2 and a half years old... :)


I am 100% with you and anyone else on that D3200 is a hot steamy pile of blur. After testing the one unit I had available, I wouldn't recommend it to anyone other than casual photographers who downsample their photos for web-use. Certainly wouldn't recommend it for photogrammetry.

But D7100 is a different story, the proof is in the pics for anyone who will take the time to do some pixel peeping. It does not have the same soft focus issues that D3200 has. It actually produces more detailed photos than 600D.

I am not biased towards Nikon either, I am actually sticking to Canons for my build (2,5x cheaper) :)


Not sure if you guys missed my photo samples from D7100 vs 600D vs D3200 - clearly Nikon D7100 captured more detail than Canon... Not just in this comparison but in my everyday use as well.

So either D7100 kit is better than 600D kit, or my Canon gear needs cleaning/fine-tuning focus :)

Also, too bad you can buy two and a half 600D for the price of one D7100 :)


A while ago, when I studied resolution charts at dxomark website, high f-stop seemed to be lens equalizer - pro lens would peform infinitely better at f4-8, but light diffraction at f11 and up seemed to degrade image quality to comparable levels with cheap lens.

Can anyone share their experience, or better yet, share comparison pics of cheap and good glass at higher fstop?

Pages: 1 2 3 [4] 5 6