The Predict Engine simulation can be saved in two different formats :
LDR (the post-processed simulation, in a PNG file),
HDR (the post-processed simulation, in an EXR file, see section bellow for more details).
Using the engine overlay, the simulation can be saved via the "PredictSuite/Engine/Game Overlay/Save Simulation" menu or using the Save button in the Interactive Settings window.
Using the engine view, the simulation can be saved via the Save button in the view.
You can also save the Predict Engine simulation by script or by defining shortcuts.
The false color and the white balance post processes can only be saved in the LDR simulation (PNG file). When the post processed output is in false colors, the false color scale can be saved with the simulation :
Using the engine overlay, the scale is saved with the simulation if the "Display Scale" option is selected on the optical instrument component (section Post Pipeline/False Colors),
Using the engine view, the scale is saved with the simulation if the option "Export false color scale with renders" is selected in the UVR preferences, section Engine Interface.
The HDR output of the Predict Engine render is stored inside an OpenExr image file (.exr). This format is specific to images that contains multiple layers (processed output, raw channels, normals, depth,...). The list of computed layers for an optical instrument can be specified on the optical instrument component when the Expert mode is enabled.
Each layer computed by the optical instrument is saved in an OpenExr layer with the name <LayerName>.<ChannelName>. For instance: "S0.R" indicates the channel R of the layer S0 and "normal.X" indicate the x component of the normal vector. The channel name is omitted when the layer has only a single channel (material IDs, Depth,...).
You can save Predict Engine simulations for different configurations of the scene and a set of camera transforms using the Batch Renderer window. Via this interface, you can select a camera, define a set of transforms for it, and create different scene configurations with a combination of pickers from the scene.
The Batch Renderer is available via the menu "PredictSuite/Engine/Batch Renderer".
The Batch Renderer window is made of a settings panel on the left and a preview panel on the right.
The settings are edited step by step :
1 : The camera or optical instrument and its transforms,
2 : The configurations,
3 : The file name format,
4 : The outputs format and resolution,
5 : The completion criteria,
A final step summarizes all the frames that will be rendered before the batch renderer is started.
The first step defines which optical instrument in the scene must be rendered and in what position(s). All Cameras and Physical Sensors can be rendered, even if they are disabled in the Hierarchy.
The scene can be rendered for a set of Transforms. The Transforms can be defined by :
A Single Position : only one Transform is rendered, the Transform of the selected optical instrument,
A Transform Picker : the scene will be rendered for each Transform listed in the picker (see the Pickers section for more details). It is also possible to compute Transforms linearly interpolated between the given Transforms to create animation frames,
An Animation Clip : the scene will be rendered for each frame of the clip (only the Transform properties of the clip will be evaluated). A Preview button enables you to preview the animation in the scene view,
A Spinner : the optical component will be placed on a sphere defined by its origin O and its radius (or distance to the origin). The inclination (theta) and azimuth (phi) of the position P are regularly spaced in the given ranges, every given step. In the example here against the Transform will be computed with theta in {50,70,90} and phi in {0,180,360}
https://www.researchgate.net/figure/Azimuthal-and-polar-spherical-coordinates-ph-and-th-of-a-point-P-on-the-spherical-surface_fig1_323551857
The second step defines the different scene configurations that should be rendered.
One configuration is defined by a combination of different Pickers from the scene (see the Pickers section for more details). If one or several Setup Pickers are listed, some elements in the scene will be enabled/disabled depending on the configuration. If other pickers are disabled because of this, they won't be used in the configurations they are not enabled in.
The third step defines how the output simulations should be named. Each file name is the combination of different parameters. The parameters you can combine are :
Scene : the scene name,
Camera : the selected optical instrument name,
Camera Transform ID : the Transform ID, each Transform has its own unique ID,
Setup ID : the configuration ID, each configuration has its own unique ID,
Frame ID : the frame ID, each combination {Transform + Configuration} has its own unique ID,
Theta/Phi : when the Transforms are defined using a Spinner, the inclination and azimuth,
Clip Name : when the Transforms are defined using an Animation Clip, the name of the clip,
{Picker} ({Picker type}) : for this picker, the name of the selected option (name of the material, name of the GameObject, name of the skybox,... depending on the picker type).
Each frame file name must be unique. As long as you have not selected a combination of elements that can differentiate each frame, the warning here against will be displayed and you will not be able to proceed to the next step.
A file name will be unique if it contains the frame ID or enough parameters to differentiate the configuration (configuration ID, pickers options) and the Transform (Transform ID, theta/phi, Camera Transform picker options) for this frame.
The fourth step defines the format of the output simulations and their resolution. The available formats are the same as when exporting the simulation from Unity (see section above).
The LDR processed frames (PNG files) can also be compiled to a movie.
The Predict Engine is a path tracer, this algorithm computes the illuminance incoming at every single point on the surface of the objects of a 3D scene. This calculation takes time and will be more and more accurate, less and less noisy, with an increasing amount of render time. The quality of a simulation can be defined by the amount of time the simulation was given to render, or by the average number of samples that were run for each pixel.
When computing a batch of frames, a completion criteria must be defined to know when a frame is ready and the renderer can move on to the next frame. The completion criteria is either defined in seconds (total render time) or in samples per pixel (SPP, average number of samples for each pixel).
It is also possible to save intermediate simulations : while the simulation is being rendered, intermediate frames are saved to see the evolution of the simulation. Frames can be saved every n SPP or every n seconds.
The last step summarizes all the combinations of camera Transforms and scene configurations. Each combination correspond to one frame and is named following the file name format defined in step 3 : this is the exact name of the output image file.
You can click on one frame name to preview the configuration and camera Transform in the scene view. As soon as the Batch Renderer window looses focus, the scene will be set back to its original state.
You can disable the frames you don't want to calculate.
When you are ready, click on the "Start Rendering" button, you will have to select the folder where the simulations will be saved, and the renderer will start rendering the frames one after another.
You can cancel a batch rendering at any time using the "Cancel" button.
If the Expert mode is enabled, you can use the Batch Renderer window to export Predict Engine standalone files (OSF and ORS) for each frame and with the required assets for all configurations to a given folder.
This can be usefull to later render the scenes on a GPU cluster for instance.