Notice: This page will be redirected to the Epic Developer Community. Spend a few minutes to get familiar with the new community-led support today.

Incomplete model The model is very noisy

Comments

23 comments

  • Avatar
    yeyuchuan 1

  • Avatar
    Ondrej Trhan CR

    Hi yeyuchuan 1,

    were the images captured in one moment?

  • Avatar
    yeyuchuan 1

    你好

    Ondrej Trhan CR

    是的 
    相机 ISO100
     
  • Avatar
    Ondrej Trhan CR

    What were your alignment and reconstruction settings? Is it possible to share your data with us?

  • Avatar
    yeyuchuan 1

    How I share data with you

  • Avatar
    Frank Zwick 1

    hello yeyuchuan 1,

    try 35mm focal length, f11 and 1/50s and optimise the position of cameras in space.

    Ìt needs a lot of tweaking to find the best settings...and of course photogrammetry friendly clothing!

    But it is possible i have done some FullBodyScans using 64 cameras:

    https://skfb.ly/6Xoxo 

    happy testing

    Frank

  • Avatar
    Ondrej Trhan CR

    You can use Wetransfer and send me the link to download it

  • Avatar
    yeyuchuan 1

    https://we.tl/t-eTGZBeABDv

    Photos and saved files are in the RAR

  • Avatar
    yeyuchuan 1

    Hello Frank Zwick 1

    The lighting environment of my scene is not bright enough, and the photos taken are all black

  • Avatar
    Frank Zwick 1

    just had a closer look to your photos - some are blurry (out of focus) and/or you are moving while capturing - refocus and faster shutterspeed can fix this problem.

    Try better framing of your scanobject (Fill the Frame!)

     

  • Avatar
    yeyuchuan 1

    Frank Zwick 1

    I still need to add camerasto populate the frame?

  • Avatar
    Ondrej Trhan CR

    yeyuchuan 1, as Frank wrote, there are some blurry images, also the jacket is not totally ideal for features creation. And the hand is just on few images full captured, so that is why it is not modeled. 

  • Avatar
    yeyuchuan 1

    Frank Zwick 1

    are you still there

    I tried using 35mm focal length

    not fully usable

  • Avatar
    yeyuchuan 1

    Ondrej Trhan CR

    hello 

    I tried the Frank Zwick 1 suggestions, but the model is still missing.

  • Avatar
    Ondrej Trhan CR

    Hi yeyuchuan 1, there are not ideal textures for the model. Did you use there some kind of texture reprojection? It could help to create better mesh. Also, with the hand, that could be problematic, as there could be just few features which are used for model creation.

  • Avatar
    Frank Zwick 1

    Hi yeyuchuan 1,

    yes - you need more light! Strong soft diffuse lighting all around...

    Try a simpler pose to start with, not a T-Pose because you will run into problems with depth of field.

    Don`t wear monochromatic featureless clothing or like Ondrej Trhan suggested use pattern-projectors to get features onto it.

    Frank

  • Avatar
    yeyuchuan 1

    Hello
    I want to figure out a problem, I took nearly 50 sets of photos, the noise of the model is still very bad, how should I solve this situation?

  • Avatar
    yeyuchuan 1

    And I watched the tutorial video, what is the function of the projector?

    I downloaded www.pi3dscan.com Before their model came out, I could see the outline of the model quite completely.

    The sharpness difference between the SLR camera and the camera should not be that big, and the texture effect is very bad after coming out

  • Avatar
    Ondrej Trhan CR

    The projector is creating texture over objects. This texture can be used for better camera alignment and the application is able to find more features, which can be used in alignment process.

    As you can see on the arm, which you send, it is quite shiny and featureless, so it is hard for the app to find there something, what can be used for alignment and modelling.

    Your model is also still wear black clothes, where is quite hard to find some features, you need something more textured.

    What you can try is XMP workflow. For this you will need some targets with coordinates or know distances between them. You will process a test model with these targets and then export XMPs for cameras, which can be used in your another projects. And the results will be in one coordinate system: https://www.youtube.com/watch?v=5VyYLaNxHz0&t=61s

    Also, the open hand is captured with small amount of images for purposes, what you want achieve. If you want to model the hand, it have to be captured from all sides on more than 15 images (the number is just a guess, but I think the hand in your dataset is correctly visible on fewer count of the images). 

  • Avatar
    yeyuchuan 1

    Ondrej Trhan CR

    The texture of the map is bad, how can I fix it?

  • Avatar
    Ondrej Trhan CR

    yeyuchuan 1, what are your texture settings? On which model is it bad?

  • Avatar
    yeyuchuan 1

    Ondrej Trhan CR

     

  • Avatar
    Ondrej Trhan CR

    Do you mean this: ?

    I think this is more a problem of model creation, as there is not enough input data to create a good model. It is also visible on the bottom side of the arm, as it looks melted.

Please sign in to leave a comment.