Maya DeepView Demo (2018)
This video shows an Autodesk Maya based tech-demo I made to test out the workflow idea of transparent depth-sliced layers being used to render a 6DoF explorable version of an immersive scene. This works in a 2D mono output, or with full stereoscopic 3D depth with adjustable parallax strength via changing the depth layer slice separation, or via Maya's stereo camera rig IPD (interpupillary distance) and convergence settings.
For my first test of a Maya DeepView workflow, the source imagery came from the Google DeepView team's dataset that used AI to compute depth from a multi-view camera rig. In the future, the same type of depth-sliced transparent layering approach for representing a volumetric scene could be created any 3D program or compositing tool.
If you textured and designed your 3D scene for use with a production renderer like Chaos Group's V-Ray Next you might use a workflow like this:
In V-Ray a transparent layered, depth sliced, volumetric scene rendering technique would be possible with a combination of Near/Far range customized camera ranges for the rendered RGBA + Z-depth channel. The second part of this workflow is the use of an animated "V-Ray Clipper" node to interactively cleave the scene at measured and precise depth zones to reveal occluded objects at progressive levels of depth. For a 49 transparent depth sliced version of the 3D scene in Maya, you would set your timeline and V-Ray Render Settings to render an animated 1-49 frame range.
The V-ray Clipper node settings would then be keyframe-animated to scan through the scene. The first frame in the timeline would have an animation key set to start the V-Ray clipper node's effect to cleave a "thin slab" of the scene with the near-distance starting at a depth distance of zero units (at the camera origin) and of a fixed thickness for the depth-slice, you need. Then on frame 49 of the animation timeline, you would expand that V-Ray clipper node's effect outwards so it only retained a sliced "slab" of the scene resting with the final slice depth sitting at the "end-wall" far depth range of the world you are inside of. The start and end frame animation keyframes on the V-Ray Clipper node would need to be set to use a "linear" animation tangent in the Maya's Graph Editor window so the depth slice rendering "slab" that animates through the scene from the near to far depth range is done in a consistent and uniform way.
Finally, you can choose to render it via a normal camera with a perspective or orthographic lens, or you could go wild and apply a raytraced panoramic lens shader like a "Domemaster" 180° angular fisheye lens, a LatLong lens, or a cube-map lens shader.
Check out an Adobe AE compatible DeepView Example:
https://github.com/Lightfielder/DeepV...
You would then batch render the Maya scene to disk, and save out the .exr or .png image sequence to use in your "DeepView" style layered depth slice playback tool
This full process is all tunable by an artist "to taste" so you could render a microscopic tiny world with this technique like a cellular scale bio-medical animation, an indoors home/office setting, a city street scene, an outdoor landscape that extends to the full horizon.
You could even do a "fulldome" planetarium style sequence of flying through meteor fields in outer space that could literally show the scene in depth-sliced form with selective depth ranges used. You could have a high detail up-close range, a huge gap of nothing, and then a lesser detail depth-slice range far-zone that would have a space background that extends out 1 astronomical unit, or further, in sliced overall depth layers (if desired).
The depth-sliced rendering in V-Ray could happen with a flat "perspective" camera and 2.5D "imageplanes" of depth, or you could use concentric "spherical" or "cubemap" panoramic rendering lens shaders (combined with depth sliced layers) to make a true 360° rendered 6DOF multi-layered scene.
For playback on a desktop or mobile device, any .obj model compatible 3D engine with WebGL (Three.js, babylon.JS, SketchFab), or Unity, Unreal would be able to playback the scene. The 3D view displays a set of stacked, semi-transparent polygon image planes, or concentric stacked cubes, or sphere. Each mesh is textured with 1 image from your folder of depth sliced images, using an "unlit" shading model.
You can also use this type of transparent layered depth-sliced media creation in a compositing package that has 3D capabilities like After Effects, Nuke, and Fusion, or inside of Photoshop's 3D model environment.
The layered scene data shown in these demos come from the Google DeepView MPI GitHub repository:
https://augmentedperception.github.io...
And these concepts were inspired directly by the DeepView MPI Viewer 6DOF webpage demo:
https://augmentedperception.github.io...