1. 30 Aug, 2016 1 commit
  2. 29 Aug, 2016 1 commit
  3. 25 Aug, 2016 1 commit
    • David E. DeMarle's avatar
      full camera for cinema · 4fd9448d
      David E. DeMarle authored
      What this does is add a new camera model where instead of
      phi and theta tracks we have a single pose track that contains
      the 3x3 orientation matrix for a camera that spins an initial orientation.
      We also save off the initial orientation (eye at and up) for each timestep
      so we can place the camera wherever we like, and then spin it.
      Placement is setup to track a named object. Other options are
      possible too, just need to provide way to let use define them.
      
      The viewer takes the current view transform,
      manipulates it via mouse drags to compute a
      desired transform matrix, and then uses the closest direction of
      projection heuristic to pick the next one.
      
      In support of the above, added a standard way to query model type to
      help viewer out and also add a tool (tests/examine_pose.py) to
      diagnose view matrices
      
      Notes:
      Initially set the viewtransform directly but lights didn't update, so
      went back to setting eye, at, and up directly.
      
      moved phis, thetas and rolls from Qt into pv_introspect so that
      cinema can downsample toward polls. generation is nearly twice as fast
      and file storage is nearly half as large with this.
      4fd9448d