why do I get different results if Ioad some picture in 3DF ?

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • curieux
    3Dflower
    • Oct 2018
    • 6

    why do I get different results if Ioad some picture in 3DF ?

    Hi
    I have the lite version and I use a turnable plate and read the pictures in with CLOSE/DEEP

    I load 75 camera picture in zephyr . After the recontruction a see 11 photos out of 75 have been oriented . I start the some process again some pictures some configuration . After the recontruction a see 24
  • cam3d
    3Dflover
    • Sep 2017
    • 662

    #2
    Hi curieux - I replied to your help ticket but will reply here too:

    If you're seeing that you're getting different numbers of cameras orientated with the same settings in 3DF Zephyr, this suggests that your photographic dataset is weak and the non-deterministic nature of GPU processing is highlighting this. Additionally, if you're changing settings or input images you are likely to see the number of cameras oriented differ, especially with feature-poor images.

    - Your subject (plastic) has a homogenous feature-poor surface. These kinds of surfaces are very hard to construct using photogrammetry. If you are just starting out, it makes more sense to practice on a feature-rich subject first, such as a tree stump or a rock.

    - With a challenging subject such as this, there are a number of things you can try which will improve the odds of a successful outcome:

    1. Use a feature-rich base for the subject to sit on top of - (**attached reference image of Perlin noise) - This will aide the camera orientation.
    2. Use a fine powder (in this case, finely ground pepper would work) and dust the subject to add feature points for Zephyr to use as surface reconstruction reference.
    3. Shoot outside on a bright yet overcast day. You will have much more light to work with than inside, and thus your camera can capture higher quality images, with lower ISO.
    4. Additionally, your subject only takes up a small portion of the overall image. Try shooting subjects which are taller than they are wide in portrait and get in as close as possible and filling the frame of your sensor for better results.
    5. If you're not already, masking is a great way to improve results by isolating the subject from the background: https://www.3dflow.net/technology/do...for-turntable/

    We have a series of video tutorials which I highly recommend to everyone, pro or beginner: https://www.3dflow.net/technology/do...hyr-tutorials/ along with extensive documentation: https://www.3dflow.net/technology/do...documentation/

    Attached Files

    Comment

    Working...