Best practices

We recommend trying some of these steps to get the best results.

Conversion

• Check the dimensions of your physical objects. Azure Object anchors works best for objects whose smallest dimension is in the range of the recommended 1m-10m.
• Inspect your 3D model in software like MeshLab for the following details.
  • Ensure that the 3D model has a triangle mesh and that the triangles on the exterior surface face outward. That is, the vertices should be oriented to make the normals follow the right-hand rule in their orientation outwards.
  • Ensure that the 3D Model is specified with the correct scale units with respect to the physical objects. The units should be one of: Centimeters, Decimeters, Feet, Inches, Kilometers, Meters, Millimeters, Yards.
  • Confirm the nominal gravity direction that corresponds to the real world vertical orientation of the object. If the object's downward vertical/gravity is -Y, use (0, -1, 0) or (0, 0, -1) for -Z, and likewise for any other direction.
  • Make sure that the 3D model is encoded in one of the supported formats: .glb, .ply, .fbx, .obj.
• Our model conversion service could take a long time to process a large, high LOD (level-of-detail) model. For efficacy, you can preprocess your 3D model to remove the interior faces.

Detection

• The provided runtime SDK requires a user-provided search region to search for and detect the physical object(s). The search region could be a bounding box, a sphere, a view frustum, or any combination of them. To avoid a false detection, set a search region large enough to cover the object. When using the provided sample apps, start the app standing about 2 meters away from the closest surface.

• Before starting the Azure Object Anchors app on a HoloLens 2 device, remove all Holograms using the Settings app. Go to System -> Holograms -> then select Remove all holograms to start with a fresh map.

  Clearing the holograms ensures objects can be properly detected in their current positions in case they were recently moved.

• After removing the holograms and before starting the app, scan the object by walking around it from 1-2 meters away a few times wearing the HoloLens.

  Pre-scanning an object and environment can help cleanup any residual surfaces created from earlier objects and scans. Otherwise, the app may see ghost surfaces leading to inaccurate alignment of your 3D model and the associated holograms. Pre-scanning the object can also greatly reduce the Azure Object Anchors detection latency, say, from 30 seconds to 5 seconds.

• For dark and highly reflective objects, you may have to scan the object at a closer range from multiple angles and multiple distances.

• The provided runtime SDK provides a few parameters to allow users to fine-tune the detection, as demonstrated in our sample apps. The default parameters work well for most objects. If you find that you need to adjust them for specific objects, here are some recommendations:

  • Use a lower surface coverage threshold if the physical object is large, dark, or shiny.
  • Allow a small scale change (for example, 0.1) for large object like a car.
  • Allow some deviation in degrees between object's local vertical direction and gravity when object is on a slope.

Next steps

In this guide, you learned some best practices to get the best results when using Azure Object Anchors to detect an object. Here are some related articles:

Troubleshooting object detection