Table of Contents

Mega real-world mesh troubleshooting guide

In the development of the Mega application, the placement of virtual content is achieved through Mesh. The visual effects and accuracy of the Mesh model can affect the placement accuracy of virtual content and even the final Mega positioning results. This document will explain some potential visual and accuracy issues with Mesh models, helping you correctly identify model defects and distinguish serious problems for feedback.

What is the mesh model

Mega Mesh is a 3D geometric model reconstructed from collected environmental data. Its primary function is to provide a real-world physical reference for virtual content, enabling virtual objects to be correctly occluded by real-world surfaces such as walls, floors, tables, and chairs, thereby achieving a more realistic mixed-reality experience.

The Mega Mesh model is automatically reconstructed from captured videos through cloud-based algorithms, not manually refined modeling, so it naturally has some degree of imperfections.

Correct expectations:

  • The overall color is generally consistent with the real environment, with large surfaces like walls, floors, and main objects having accurate colors.
  • The textures of major structures are recognizable (e.g., floor tile patterns, wallpaper designs, exhibit appearances).
  • The model provides good lighting and color blending effects in AR.

Minor imperfections are completely normal, including:

  • In areas with complex lighting, texture colors may appear abnormally dark, bright, or discolored.
  • Distant objects or edge areas may have slightly blurred or stretched textures.
  • Mega provides an LOD model, so there may be slight color inconsistencies between LOD blocks.
  • At stitching points between different perspectives, there may be minor texture misalignments or seams.
  • Small details like text or signs may not be clear.
  • Dynamic objects like people or cars may leave slight ghosting or color smudges.
  • Transparent or reflective surfaces like glass or water may have missing structures or depressions.
  • Walls or floors may not be perfectly flat, with minor bumps or "burrs."
  • Small holes may exist in the model.

These imperfections typically do not significantly impact the experience and are reasonable limitations of automatic reconstruction algorithms, requiring no special handling.

Which model anomalies need to be addressed

The following situations are serious defects that will directly affect the application's effectiveness and require attention and handling.

Issue type Typical manifestation Possible impact
Large-area missing or voids Non-reflective, transparent entire wall completely lacks Mesh, virtual objects cannot be correctly occluded Inability to achieve correct occlusion in this area
Large-area structural errors An entire wall is incorrectly reconstructed as slanted, or the room's dimensions severely mismatch reality Causes virtual objects to misalign with the real environment, severely affecting occlusion
Severe distortion or tilting Obvious ground tilt, wall twisting, severe object misplacement Virtual content misalignment, occlusion anomalies
Severe noise or artifacts Large amounts of floating debris in the air Occlusion flickering, performance degradation
Critical areas completely wrong Key interaction areas' Mesh completely incorrect (e.g., stairs, display stands) User interaction failure, experience ruined
Model layering Seeing ground or walls layered at the same location Occlusion errors, incorrect positioning
Large-area color completely wrong Walls, floors, etc., exhibit strange colors or entirely incorrect colors Unrealistic overall environment, visual dissonance

Issue handling and feedback

After encountering the above serious model defects, please collect information and provide feedback in the following ways to help us quickly locate and fix the issues.

Collect necessary information

  1. Screenshots or videos

    • Screenshots: At least 3 from different angles, with the problematic areas marked.
    • Videos: Record a 10–30-second video using a phone or screen-recording software, focusing on the abnormal Mesh area.
  2. Problem description

    When reporting an issue, provide the following information:

    • Scope of the issue: Whether the problem affects the entire map or a specific local area.
    • Map type: Mega offers several mapping methods; specify which type is problematic.
    • Problem location/area: The exact location of the issue, e.g., "the central display area in the first-floor hall of the museum" in a single map, or in a large-scale map, specify which sub-maps have fusion issues like layering.
    • Problem description: Combine with screenshots or videos to detail the problematic area and the anomaly. For multi-maps, clarify the sub-map names. For example, "In the XX map, the 3rd floor of the office building, the wall model on the west corridor is tilted by about 15 degrees," or "Layering occurs between the west side of sub-map A and the east side of sub-map B."

    Avoid vague statements like "the model has issues" or "the Mesh doesn’t look good." Lack of screenshots and specific descriptions will significantly delay issue resolution.

  3. Map information

    • Mapping report: Provide the mapping report of the problematic map from Mega Block.

      Download mapping report

    • Localization service information: Export the localization service information in Unity.

      Localization service information

Initiate feedback

Send screenshots or screen recordings, problem descriptions, and map information to EasyAR staff. After analysis, they will provide feedback on the solution.

  • Process: We will review the original collected data and cloud reconstruction results based on the materials you provide.
  • Possible causes: It could be due to environmental changes during collection, issues with the collection device, or limitations of the mapping algorithm in specific scenarios.
  • If collection was not performed according to specifications: We will assess whether the area needs to be recollected or supplemented, and then remap the data.
  • If the collected data is normal: There may be shortcomings in the cloud reconstruction method. We will iteratively optimize it in future versions and provide you with a solution.

Understanding the functionality and limitations of the Mesh model, distinguishing between "normal flaws" and "real issues," and providing feedback as required are key to efficient problem-solving.