Loading ...
Sorry, an error occurred while loading the content.
 

Re: 5 cameras with a plugin

Expand Messages
  • Eric Knisley
    It s a trade-off. The fisheye method is slower, but virtually guarantees that all effects will render correctly, and can be more accurate. The multiple camera
    Message 1 of 9 , Nov 23, 2009
      It's a trade-off. The fisheye method is slower, but virtually guarantees
      that all effects will render correctly, and can be more accurate. The
      multiple camera method can be much faster (since it does not absolutely
      require ray-tracing every pixel), but can introduce errors with
      post-process effects like motion blurs, which often use the edge of the
      frame and camera vector as parameters.

      In my experience, it's best to be at least familiar with *both*
      approaches, and to make your production decisions based on that
      familiarity. Sometimes one way is better/faster, sometimes another.

      --ek

      Planetarium Production wrote:

      > Thank you very much for this plug. Usually I don't use 5 cameras but a
      > fish eye camera w/mental ray, vray or final render. Are you sure to get
      > better results with 5 cams ? I never saw bad things with a fisheye cam
      > in PNG or TGA. What is your opinion Dietmar or someone else ?

      --
      Eric Knisley
      Cartoonist | Illustrator | Animator | Photographer
      Carrboro | NC | US | www.silent-k.net
      +1 919-491-7406
    • Dietmar
      Hi Cyril, To be honest, we do not have experience with fisheye cams within 3ds max. I believe it works just fine. But i think the 5-cam-rig had some advantages
      Message 2 of 9 , Nov 23, 2009
        Hi Cyril,

        To be honest, we do not have experience with fisheye cams within
        3ds max. I believe it works just fine. But i think the 5-cam-rig
        had some advantages (in our particular case):

        - Rendering can be distributed to more "small" machines
        - I liked it, to have an impression while working within max
        what in fact will happen in your front, left, right etc.
        This was important for us, since we used live action footage
        of the character which was motion-tracked and we needed
        the front-view to get as close to the original footage as
        possible. As far as i have tried (eg vray domecam) you dont
        get a proper disorted image in your viewport (you may prove
        me wrong)

        Of course it has a couple of disadvantages (no post effects,
        glom supports no alpha so you have to stich a separate alpha
        pass etc.)

        ... but of some of them "The Didder" takes care :)

        So how is your fisheye w/mental ray set up?

        Best wishes,

        Dietmar

        Planetarium Production wrote:

        > Thank you very much for this plug. Usually I don't use 5 cameras but a
        > fish eye camera w/mental ray, vray or final render. Are you sure to get
        > better results with 5 cams ? I never saw bad things with a fisheye cam
        > in PNG or TGA. What is your opinion Dietmar or someone else ?
      • Tom Casey
        I totally agree with this comment. Just use what works for you for any particular scene. It s our approach. The real goal is to get a visually effective
        Message 3 of 9 , Nov 23, 2009
          I totally agree with this comment. Just use what works for you for
          any particular scene. It's our approach. The real goal is to get a
          visually effective image up on the dome. No one ever asked
          Michelangelo what paintbrush he used to paint the Sistine ceiling.
          We just enjoy the finished work.

          Tom

          On Nov 23, 2009, at 9:04 AM, Eric Knisley wrote:

          In my experience, it's best to be at least familiar with *both*
          approaches, and to make your production decisions based on that
          familiarity. Sometimes one way is better/faster, sometimes another.

          ************************************************
          H o m e R u n P i c t u r e s

          Tom Casey
          President & Creative Director

          100 First Avenue - Suite 450
          Pittsburgh, PA 15222
          412-391-8200
          mailto:tom@...
          http://www.homerunpictures.com
        • Jon Strawn
          Just a note about stitching, there are other 5-camera stitching programs available that DO support alpha channels. During production on Tales of the Maya
          Message 4 of 9 , Nov 23, 2009
            Just a note about stitching, there are other 5-camera stitching programs available that DO support alpha channels. During production on "Tales of the Maya Skies," we used the Sky-Skan stitcher extensively and with great success.

            -jon
            ARTS Lab at the Universtiy of New Mexico
            artslab.unm.edu
            www.mayaskies.org
            www.domefest.org
          • pauldavidbourke
            ... In case anyone is wondering why there can be a performance difference between fisheye and cubic maps ... one reason is that a clever renderer (raytracer or
            Message 5 of 9 , Nov 24, 2009
              > Thank you very much for this plug. Usually I don't use 5 cameras but a fish eye camera w/mental ray, vray or final render. Are you sure to get better results with 5 cams ? I never saw bad things with a fisheye cam in PNG or TGA.

              In case anyone is wondering why there can be a performance difference between fisheye and cubic maps ... one reason is that a clever renderer (raytracer or not) can be more efficient with a narrower field of view ... more geometry can be culled for primary rays for example but there are more cunning tricks available as well, not that package xyz will necessarily exploit them.

              My choice is often to use cubic maps because then I can more easily repurpose the content for different dome orientations rather than rerender. Indeed I believe the industry and the standards group has gone down the wrong path with regard to fisheye frames as the "master" format for playback ... but that's another story.
            • Tom Casey
              The two approaches have other differences as well. Although the placement of pixels in finished frames of stitched or fisheye rendered can end up being
              Message 6 of 9 , Nov 24, 2009
                The two approaches have other differences as well. Although the
                placement of pixels in finished frames of stitched or fisheye
                rendered can end up being identical (meaning everything in the scene
                appears in the same location), there are some quality issues. Since
                a stitcher needs to distort the square source images to create the
                circle, some "approximations" occur where the pixels need to be
                "recalculated." This is called sampling. Some stitchers are better
                than others at this. With a fisheye render, since the renderer is
                originally creating the pixels in the circle format, they are what
                they are, so the quality can be better.

                In reality, this only becomes apparent in certain scenes... a
                starfield, for instance, can suffer greatly from the stitching approach.

                Tom


                On Nov 24, 2009, at 3:55 AM, pauldavidbourke wrote:

                In case anyone is wondering why there can be a performance difference
                between fisheye and cubic map

                ************************************************
                H o m e R u n P i c t u r e s

                Tom Casey
                President & Creative Director

                100 First Avenue - Suite 450
                Pittsburgh, PA 15222
                412-391-8200
                mailto:tom@...
                http://www.homerunpictures.com
              • pauldavidbourke
                ... Mine uses straightforward supersampling so final quality is dependent on the processing time you can afford .... I d be surprised (but then I often am) if
                Message 7 of 9 , Nov 24, 2009
                  > The two approaches have other differences as well. Although the
                  > placement of pixels in finished frames of stitched or fisheye
                  > rendered can end up being identical (meaning everything in the scene
                  > appears in the same location), there are some quality issues. Since
                  > a stitcher needs to distort the square source images to create the
                  > circle, some "approximations" occur where the pixels need to be
                  > "recalculated." This is called sampling. Some stitchers are better
                  > than others at this.

                  Mine uses straightforward supersampling so final quality is dependent on the processing time you can afford .... I'd be surprised (but then I often am) if converting between appropriate size cubic maps and fisheye with supersampling of 3 causes discernible degradation. I accept that errors can easily occur when converting between inappropriate sized cubic maps, this works both ways ... going from cubic maps of too low resolution but also cubic maps with too high a resolution.

                  > With a fisheye render, since the renderer is
                  > originally creating the pixels in the circle format, they are what
                  > they are, so the quality can be better.

                  But remember that most projection systems are warping the fisheye frames anyway, so you are ending up with a resampling anyway. The final comment in my last email was suggesting that perhaps we have gone down the wrong path of using fisheyes for playback. If rather we settled on cubic maps as the "master format" just think how much power that would give, interaction during movie playback for example. This is something I almost always use in the iDome, it is a powerful capability. But even without interaction, it would allow one movie (a movie of cubic maps) to be used in any orientated dome without modification.
                • Tom Casey
                  If you think about the process we go through to get an image up on the dome, you have to laugh... render, stitch to circle, slice to provide the specific
                  Message 8 of 9 , Nov 25, 2009
                    If you think about the process we go through to get an image up on
                    the dome, you have to laugh... render, stitch to circle, slice to
                    provide the specific projector arrangement, or distort for mirror
                    projection... yes, there had to have been a better approach.

                    Tom


                    On Nov 25, 2009, at 1:29 AM, pauldavidbourke wrote:

                    But remember that most projection systems are warping the fisheye
                    frames, so you are ending up with a resampling anyway.

                    ************************************************
                    H o m e R u n P i c t u r e s

                    Tom Casey
                    President & Creative Director

                    100 First Avenue - Suite 450
                    Pittsburgh, PA 15222
                    412-391-8200
                    mailto:tom@...
                    http://www.homerunpictures.com
                  Your message has been successfully submitted and would be delivered to recipients shortly.