• transient: First steps with mitransient (+Mitsuba 3): creating a scene, editing the scene and visualizing the transient result. There are more advanced tutorials for transient volumetric rendering, frequency space rendering and advanced visualization.
• transient-nlos: Example on how to render NLOS simulations from the Python interface. For easier setups you can also check tal, a Python library with a shell interface that simplifies this process.
• polarization: Most of the tutorials do not simulate the polarization of the light. This folder contains multiple samples for how to simulate time-resolved polarization of light, and how to visualize the result/Stokes vectors.
• diff-transient: Differentiable transient rendering. We show examples for gradient-based optimization with a transient signal (backward-mode autodiff) and forward inverse rendering (forward-mode autodiff) in the time domain.
• angulararea-emitter: Examples for our angulararea plugin, which acts as an area light that emits within a restricted angular range.

To create a scene and simulate a NLOS setup, you will need to set up the following components:

From here you have two alternatives:

• Configure the XML/Python code of the scene youself, using the help below
• Use https://github.com/diegoroyo/tal. tal is a toolkit that allows you to create and simulate NLOS scenes with an easier shell interface instead of directly from XML/Python.

Here's a list of the required plugins and configuration in your scene:

• Relay wall: The relay wall must be defined using a rectangle shape with ID "relay_wall" i.e. <shape type="rectangle" id="relay_wall">
  • Camera: Inside this shape, you must place a nlos_capture_meter plugin, which will focus your camera towards a set of points in the relay wall. In order to configure the nlos_capture_meter, read its documentation.
• Laser/light source: You must define an active illumination emitter that has the ID "laser". You can use the existing projector plugin with a low enough fov parameter. In order to point the laser to the relay wall, we recommend using our helper functions instead of using the transform plugin. In our example, we only define the translate part of the transform:

  <emitter type="projector" id="laser">
      <transform name="to_world">
          <translate x="-0.5" y="0.0" z="0.25"/>
      </transform>
      <rgb name="irradiance" value="1.0, 1.0, 1.0"/>
      <float name="fov" value="0.2"/>
  </emitter>

  • Pointing the laser source to the relay wall: You can use our helper functions for this purpose. We offer three alternatives:
    • mitransient.nlos.focus_emitter_at_relay_wall_3dpoint(target, relay_wall, emitter): Accepts a mi.Point3f with the XYZ coordinates on the relay wall where the laser should be pointed at.
    • mitransient.nlos.focus_emitter_at_relay_wall_uv(uv, relay_wall, emitter): Accepts a mi.Point2f with the UV coordinates of the shape that you used as relay wall (e.g. if your relay wall is a rectangle, passing mi.Point2f(0.5, 0.5) will point the laser at the center)
    • mitransient.nlos.focus_emitter_at_relay_wall_pixel(pixel, relay_wall, emitter): Accepts a mi.Point2f with the pixel coordinates (e.g. if you capture a grid of 3x3 points with your nlos_capture_meter, passing mi.Point2f(1, 1) will point the laser at the center of the center pixel)
    • Note that you need to pass the relay_wall and emitter parameters, which should be the shapes with ID "relay_wall" and "laser". See the examples Jupyter notebook for more information.
• Hidden object and occluders: You can add other shapes in the locations that you want with no restrictions.
• Integrator: We provide a transient_nlos_path plugin which implements a path-tracing algorithm with specific techniques which work better in NLOS setups. We strongly recommend that you read the documentation and our paper for further information.