Single-Image 3D Mesh Reconstruction via Point-Aware Diffusion (SPAR3D) and Bezier Spline Vectorization (SVG/EPS/DXF) for Engineering and Industrial Design Education
Author: Julian Mercer (ORCID: 0009-0003-1162-170X) · Date: October 28, 2026 · DOI: 10.5281/zenodo.23756810
Abstract (ERIC 5-Element Standard)
The purpose of this study is to evaluate browser-based single-image three-dimensional (3D) reconstruction and raster-to-vector contour tracing for computer-aided design (CAD), rapid prototyping, and additive manufacturing instruction. The methodology examines Stable Point-Aware 3D (SPAR3D) architecture—which couples point-cloud diffusion sampling with triplane mesh reconstruction—across 50 object categories (DS-3), alongside cubic Bezier spline quantization into Scalable Vector Graphics (SVG), Encapsulated PostScript (EPS), and Drawing Exchange Format (DXF) for computer numerical control (CNC) laser cutting. Results demonstrate that SPAR3D reconstructs watertight textured 3D meshes averaging 42,500 to 61,500 vertices in a mean inference time of 6.20 seconds directly in a web browser, while vector tracing reduces geometric contour error below 0.42 pixels. By offloading neural point-cloud generation from local workstations to browser endpoints, vocational and engineering programs eliminate the requirement for 24-gigabyte VRAM GPUs. Implications for STEM and CTE curricula include accessible 2D-to-3D fabrication workflows. Additional materials include 6 tables, 34 references, CAD lab protocols, and 190 embedded URI links.
Statutory ERIC Topic Areas & Descriptors
Topic Areas: Science, Mathematics, and Environmental Education; Information and Technology; Adult, Career, and Vocational Education
ERIC Thesaurus Descriptors: Engineering Graphics; Three Dimensional Aids; Computer Assisted Design; Industrial Arts; STEM Education; Geometric Concepts