Reconstruction settings

In each place where reconstructions can be made, it is also possible to configure the reconstruction settings. This documentation page describes each available setting.

There are two ways of configuring reconstruction settings: Either the settings can be configured individually, or a setting preset can be selected which determines the value of all settings. The ‘+’ button to the right of the setting preset selector creates a new preset from the currently configured individual settings.

Using a preset is useful to apply the same settings to many reconstructions. It may also be used to ensure that all settings are set to known values.

In the sections below, the available individual settings are listed. Setting presets contain some additional (technical) settings which are not available as individual settings. The available settings for presets are listed on a separate documentation page.

Mesh settings

Mesh resolution [mm]

Defines the mesh resolution in millimeters. Lower is better, allowing to capture smaller details.

Note that lowering the mesh resolution increases the GPU memory requirements. If you encounter memory-related issues, please try increasing the resolution.

Mesh triangle limit

Defines the maximum triangle count for reconstructed meshes.

Mesh simplification (%)

Defines how strongly meshes are simplified compared to their original triangle count. For example, a setting of 20% means that meshes are simplified to 20% of their original triangle count (or to the configured mesh triangle limit, if that is lower).

Mesh smoothing

Defines how strongly meshes are smoothed. Smoothing can be useful to remove small spikes. However, it can also cause edges and corners to become too smooth.

Mesh quality

Defines the quality of reconstructed meshes. By default, the program uses a fast mesh simplification algorithm that will relatively often create degenerate meshes. This has little impact for display purposes, but will make these meshes harder to edit. In contrast, the high-quality setting uses an algorithm that will output higher-quality meshes, but which takes more time during processing.

Texture settings

Texture resolution

Defines the resolution for the texture map. Larger resolutions can store sharper and more detailed appearance if the cameras provide sufficiently detailed raw video. However, they are slower to process while requiring more memory, and the result will take more bandwidth for video storage and transmission.

Note that increasing the texture resolution increases the GPU memory requirements. If you encounter memory-related issues, please try lowering the resolution.

Texture bleed prevention intensity

Defines the strength of texture bleed prevention. Larger values will prevent texture bleed (foreground colors incorrectly appearing on background surfaces, or the other way round) more effectively, but at the same time increase the chance of missing textures.

Texture encoding

Defines the encoding quality of reconstructed textures. By default, the program encodes textures as a video. This achieves efficient storage and playback. However, the encoding is lossy. If you want to preserve the highest possible quality, for example if you want to export the reconstructed meshes and continue editing them in an external program, then choose lossless encoding instead. Please note that lossless encoding will result in large file sizes, and such files are incompatible with ScannedReality’s playback plugins.

Other settings

Transparent back side

Sets whether the program will reconstruct the back side of the subject as transparent.

We recommend to enable this for front-facing camera configurations which do not observe the back side of the subject.

Video stabilization

Sets whether the program will use the same mesh geometry for subsequences of video frames, giving the video temporal consistency and allowing for efficient compression.

We recommend to only set this to off in case you either have a specific use-case that requires maximizing per-frame geometrical accuracy, if you apply video stabilization afterwards in an external program, or potentially for offline rendering.

Depth bias compensation [mm]

Time-of-flight cameras like the Azure Kinect and Femto Bolt have a material-dependent bias in their depth measurements. In particular, for materials like skin, the infrared light sent out by the cameras reflects not only on the material surface, but also below the surface, such that the measured depth is larger than the actual depth. To correct this, set the depth bias compensation to a positive value in order to reduce the measured depth values.

By default, this setting is set to 10 mm. We have found 8 mm to work well with Azure Kinect cameras and 15 mm with Femto Bolt cameras. You may need to tune this for your setup.