How Mesh Regression Boosts 3D Cloud Design Testing Efficiency

Why Mesh Regression?

With the development of 3D cloud design tools, users can design interiors, public spaces, architecture, and real estate entirely in the cloud. These tools contain many modules (hard‑decoration, customization, BIM), generating large testing workloads that require a targeted regression solution.

What Is Mesh?

A mesh is a unified representation of a 3D model where all surfaces are split into triangles.

Application Scenarios

BIM plumbing and electrical pipes

Custom door and window shapes

Custom editor sweep and slant sweep

Hard‑decoration tiles, floors, water‑jet cuts, stone, trim, integrated wall panels

Design and Implementation

4.1 Know the Business

First review the design ideas of the business to obtain necessary information.

4.2 View Development Code

Since the middle platform only stores data without decompression and conversion logic, custom decompression must be written.

4.3 Obtain Source Information

V represents a triangle vertex; each line contains three XYZ coordinates, and every three lines form a triangle.

4.4 Validation Calculation

With triangle vertices, compute area using the Heron formula:

S = (x1*y2 - x1*y3 + x2*y3 - x2*y1 + x3*y1 - x2*y2).

After mesh processing, the sum of all triangle areas should equal the wall area in the tool.

4.5 Types of Bugs Solved

4.5.1 Missing Faces

Problem: generated roofs have missing faces.

Solution: verify by summing triangle areas.

4.5.2 Water‑Jet Shape Missing

Problem: textures and shapes missing on water‑jet surfaces.

Solution: validate by counting mesh triangles (area check is inefficient due to tiny triangles).

4.5.3 Incorrect Normals

Problem: mesh orientation reversed, causing preview errors.

Solution: validate mesh normal vectors.

4.6 Interface Integration

Integrate the three validation methods into the API platform so that underlying logic automatically compares actual and expected data, achieving automated regression.

4.7 Results

After running for a while, the three validation methods discovered several bugs.

Automation increased bug detection from 0% to 13%.

Performance Advancement

5.1 UV Issues

Even after fixing previous bugs, UV mapping problems remain, which are hard to automate because they involve texture retrieval, UV coordinate validation, and matrix conversion.

5.2 Solution Idea

Mesh textures are controlled by an MTL file. By knowing the original image and applying the company’s specific ratio, UV coordinates can be embedded into the mesh to achieve the expected preview.

5.3 Design Concept

Using three.js to render and compare effects, the design can be visualized and compared, improving efficiency.

5.4 Result Display

Extensive development work produced the following visual results.

Planning and Outlook

Increase Scene Coverage

The hard‑decoration tool line already supports many custom parametric models; future work will extend full support to custom tools.