Announcement

Collapse
No announcement yet.

Capturing an egg

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Capturing an egg

    In a perfect world I could put a couple registration stickers on an egg and take photos of it on a white background, turning it around to get all the surface information.

    But, I think to use Zeyphyr, I need to suspend the egg from a wire and take photos of it in the center of a room with lots of background details?

    Any ideas of the simplest way to do this?

  • #2
    Hi Gazaah,

    there is no need for registration stickers or for complex hanging threads. Simply put your egg on something with high contrast (i.e. a white or green paper - depending on the color of the egg -, to speed up the masking process) take some photos, rotate the egg, take more photos, mask out the unwanted parts and you're set


    You can follow this tutorial

    https://www.youtube.com/watch?v=dGRw8LbXknU

    I guess it would be easier to simply take photos by moving around the egg since it may roll if you try to move it, but it shouldn't be an issue at all.

    Comment


    • #3
      Success!!

      Andre, your software is awesome. After only a few days of learning we have good enough results to share with others and I know we will do better as we master masking and photography. We could not get this to work in Autocad's Recap Photo and it only took a few runs to learn Zephyr, I am impressed. I was actually a beta tester for the Autocad software, so I thought I could get it to work as I had successfully used it in beta... but their automation has a LOT of trouble with round objects.

      Because we want to work with spotted eggs, the Zephyr software had trouble tracking speckles which may not have unique edges for easy detection, the software would jump between tracking very similar speckles, creating bumps in the mesh. We did end up using 4 tiny, colored registration bumps. The registration bumps plus using a few unique edge speckles made great control points.

      When we used about 400 images, pulled from HD video, the models started coming out nicely without masking, we used 2% difference between images.

      I will be trying some maximal contrast backdrops, such as 100% saturated bright magenta, and see if that allows us to better use masking. Our bluescreen wasn't working. Masking works very well when carefully done manually per frame, but slowed down the process without enough contrast. I believe it is hard because heavily blotched or speckled eggs have a confusing visual boundary. Even against contrasting fabric back drop, sometimes the computer's automatic detection would create the wrong edge boundary.

      I am giving a small presentation on how to use your software today to some other scientists, as they are taking aerial photos of the tree canopy but cannot tell which trees are tallest, and I think your software could be a solution with much less hassle than trying to use Autocad's cloud based system (some jobs get put in a queue and it takes a long time even begin them AND there's very little control over parameters.)

      Your pricing is also fairly reasonable compared to other top-end softwares. Pro pricing seem great. Aerial is a little high, but the educational and research pricing of $1477 is not ~too ~bad as long as it is not annual renewal.

      Typically, we buy academic licenses for research software in the U.S. for lower prices, say $250-$500 USD a year with required annual renewals, creates low initial profit for the seller but creates a longer cusomter lifespan. The free education or research licenses scare the hell out of our IT techs who are afraid that since we didn't actually buy it somehow we will get in trouble for using it. So we don't end up using free licenses for research.

      Attached is a representative photo of the sort of speckles that the software has trouble tracking, as well as what I'll call "edge detection hell" for masking. Any shadowing at all will interact with splotches (see splotchy egg photo) and the automated masking can get confused. I'll see how the magenta works!!

      -Connie


      Comment


      • #4
        Excellent and thanks for the update .
        No worries, we hate rentals or pay per use software as consumers, and we don't want to change policy. All our licenses are perpetual - including Edu - and updates after the first year are optional.
        Also, with the purchase of a single Edu license, we can generate temporary free keys for classrooms and students projects.

        Comment


        • #5
          This. Software. Is. Fantastic.

          Due to the problems with blotches and speckles being picked up as holes and boundary edges by the masking, I had to come up with a Photoshop batch edit to remove the backgrounds manually and set them to pure white #000000.

          Once I remove backgrounds in Photoshop, I can get a fantastic model from about 20 images. Below is a scan of a whooping crane egg, we will be printing these out in full 3D color to use to hide sensors. This way we can understand how whooping cranes rotate their eggs to diagnose why some cranes aren't having healthy babies.

          Comment


          • #6
            Though I will note the meshes that come out are nowhere near ready for 3D printing. I wonder if it is because my process is wrong? The scan posted above has about 2,500 flaws that prevent it from being understood as a single triangle mesh. Each of those colored balls on a stick represent a defect to repair.

            I run my meshes through Autodesk's Meshmixer before printing. The inspector found a very large number of problems with the triangle mesh.

            Comment


            • #7
              Hi gazaah,

              thank you for your kind words!

              which issues does your mesh have exactly? Which format file are you exporting to ?

              If you want to share your dataset / mesh we can have a look at it

              Comment


              • #8
                The biggest export problem is that the mesh isn't welded. The exported mesh is made of >500 distinct meshes that can interpreted as different "shells" which have distinct boundaries.

                The reason I am using Meshmixer for analysis is that it is very good as welding mesh boundaries using the "close cracks" tool. Once I use the "close cracks" tool, I edit the mesh as a single shell. Attached is the after and before images showing the unwelded boundaries being fixed.

                Comment


                • #9
                  Hi gazaah,

                  that doesn't sound right, maybe it's a meshmixer issue with some filetype? can you try changing the export format? If there are holes you can use the fill holes filter in zephyr. can you share a zep file and photos with me so i can double check?

                  Comment

                  Working...
                  X