Mega scene mesh troubleshooting guide
In Mega application development, the placement of virtual content relies on the Mesh. The visual quality and accuracy of Mesh models directly impact the precision of virtual content placement and even affect the final Mega localization results. This document explains potential visual and accuracy issues with Mesh models, helping you correctly identify model defects and report critical problems.
What is a mesh model
Mega Mesh is a three-dimensional geometric model reconstructed from collected environmental data. Its primary function is to provide real-world physical references for virtual content, enabling virtual objects to be correctly occluded by real-world surfaces like walls, floors, tables, and chairs, thereby achieving more realistic mixed-reality experiences.
Mega Mesh models are automatically reconstructed from captured videos by cloud-based algorithms, not manually fine-modeled. Therefore, they inherently contain some imperfections.
Realistic expectations:
- Overall color generally matches the real environment, with major surfaces like walls, floors, and object bodies displaying correct colors.
- Textures of major structures are recognizable (e.g., floor tile patterns, wallpaper designs, exhibit appearances).
- The model provides good lighting, shadow, and color blending effects in AR.
Minor imperfections are normal and common:
- In areas with complex lighting, textures may appear abnormally dark, bright, or discolored.
- Distant objects or edge areas may exhibit slight texture blurring or stretching.
- Mega provides LOD models, so slight texture color inconsistencies may exist between LOD blocks.
- Small areas of texture misalignment or seams may occur at viewpoint stitching points.
- Fine details like text or signs may be unclear.
- Slight ghosting or color smearing may appear from dynamic objects like people or vehicles.
- Transparent or reflective surfaces (e.g., glass, water) may have structural gaps or depressions.
- Walls or floors may not be perfectly flat, showing minor bumps or "burrs".
- Small holes may exist in the model.
These imperfections typically do not severely impact the experience and are inherent limitations of automatic reconstruction algorithms, usually requiring no special action.
Which model anomalies require attention
The following situations constitute serious defects that directly impact application effectiveness and need attention and resolution.
| Problem type | Typical manifestation | Potential impact |
|---|---|---|
| Large-area missing or holes | Entire non-reflective, non-transparent walls completely lack Mesh, preventing correct occlusion of virtual objects | Inability to achieve correct occlusion in that area |
| Large-scale structural errors | An entire wall reconstructed incorrectly as slanted, or room dimensions severely mismatched with reality | Causes misalignment of virtual objects with the real environment, severely affecting occlusion |
| Severe deformation or tilting | Noticeably slanted floors, twisted walls, objects severely misplaced | Virtual content placement disorder, abnormal occlusion |
| Severe noise or artifacts | Large amounts of floating fragments in the air | Occlusion flickering, performance degradation |
| Critical area completely wrong | Mesh completely incorrect in key interaction areas (e.g., stairs, display stands) | User interaction failure, experience disruption |
| Model layering | Seeing layered ground or walls at the same location | Incorrect occlusion, inaccurate localization |
| Large-area color completely wrong | Walls, floors, etc., displaying strange or entirely incorrect textures | Unrealistic overall environment, visual dissonance |
Problem handling and feedback
Upon encountering the serious model defects listed above, please collect information and provide feedback as follows to help us quickly locate and resolve the issue.
Collect necessary information
Screenshots or video
- Screenshots: At least 3 from different angles, clearly marking the problem area.
- Video: Record a 10-30 second video using a phone or screen recording software, focusing on the Mesh anomaly area.
Problem description
When reporting the issue, include the following information:
- Problem scope: Whether the entire map is affected or only a specific local area.
- Map type: Mega offers several mapping methods; specify which type is problematic.
- Problem location/area: The precise location of the issue. For a single map, specify like "central display stand in the museum first-floor hall". For large-scale maps, indicate specific sub-maps where layering occurs during fusion.
- Problem description: Using screenshots or video, detail the affected area and the anomaly. If multiple sub-maps are involved, name them. Examples: "The wall model in the west corridor, 3rd floor of the office building in XX map, is tilted by approximately 15 degrees" or "Layering occurs between the west side of sub-map A and the east side of sub-map B".
Do not simply state "the model has issues" or "the Mesh looks bad". Lack of screenshots and specific descriptions will significantly delay resolution.
Map information
Mapping report: Provide the mapping report for the problematic map from Mega Block.

Localization service info: Can be exported within Unity.

Initiate feedback
Send screenshots/screen recordings, the problem description, and map information to EasyAR staff. After analysis, we will provide feedback on the resolution.
- Process: Based on your materials, we will examine the original captured data and cloud reconstruction results.
- Possible causes: Could be environmental changes during capture in that area, device issues during capture, or limitations of the mapping algorithm in specific scenarios.
- If capture wasn't standard: We will assess if re-capturing or supplementary capture of that area is needed, followed by re-mapping.
- If capture data is normal: The cloud reconstruction method might have shortcomings. We will iterate and optimize in future versions and provide you with a solution.
Understanding the function and limitations of the Mesh model, distinguishing between "normal imperfections" and "real problems", and providing feedback as required are key to efficient problem resolution.