Smartphone Artifact Replicas for 3D Modeling Projects

ISEF Category: Technology Enhances the Arts

Subcategory: 3D Modeling  ·  Difficulty: Intermediate  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

A phone can do more than take pictures. With the right capture path, it can help you rebuild a tiny object in 3D, then print a copy that you can measure against the original. That makes this project feel like digital magic, but the real win is precision. You get to test how close a low-cost setup can get to a reference scan.

What Is It?

Photogrammetry turns lots of overlapping photos into a 3D model. Your phone takes the photos, a turntable rotates the object, and software finds matching points from image to image. Think of it like solving a puzzle where every photo gives you a new piece of the object’s shape.

Gaussian splatting is a newer way to represent 3D scenes using many tiny blurry points that carry color and position. A mesh is the more familiar skin of triangles that most 3D printers need. Your project tests how well a low-cost pipeline can move from photos, to a digital model, to a physical print, then compares that print to a reference scan to see where the shape drifts.

Why This Is a Good Topic

This is a strong science fair topic because you can change one part of the pipeline at a time and measure the effect on error. You can test camera angle, number of photos, lighting, background, or software choice. The project connects to cultural heritage, product design, and low-cost digital preservation. You can learn image capture, 3D reconstruction, error analysis, and how to judge whether a printed replica matches the original shape.

Research Questions

• How does the number of photos affect surface error in a smartphone photogrammetry model?
• What is the effect of lighting consistency on mesh completeness and print accuracy?
• Does a rotating turntable reduce reconstruction error compared with hand-held capture?
• To what extent does object texture influence feature matching on a ceramic figurine?
• Which capture angle range produces the lowest average deviation from a reference scan?
• How does Gaussian-splatting output compare with direct mesh reconstruction for print fidelity?

Basic Materials

• Smartphone with a good rear camera.
• Small motorized or manual turntable.
• Tripod or phone stand.
• Plain matte backdrop in a neutral color.
• Two or more desk lamps with consistent bulbs.
• A small ceramic figurine or similarly sized object.
• Ruler or caliper for basic size checks.
• Free photogrammetry software that supports mesh export.
• Access to a 3D printer, if you plan to print the result.

Advanced Materials

• Smartphone with manual camera controls.
• Calibration object or scale bars for size reference.
• Precision digital caliper.
• Desktop computer with stronger GPU support.
• Gaussian-splatting reconstruction software.
• Mesh cleanup software for repair and decimation.
• Resin 3D printer and post-processing tools.
• Reference scan from a structured-light scanner or high-resolution desktop scanner.
• Image registration or metrology software for surface deviation analysis.

Software & Tools

• Meshroom: Reconstructs 3D meshes from overlapping photos and helps you compare capture settings.
• COLMAP: Estimates camera positions and supports advanced photogrammetry workflows.
• Blender: Cleans meshes, aligns models, and prepares files for printing.
• ImageJ: Measures image sharpness, contrast, and simple calibration targets.
• Python: Automates file handling, error calculations, and comparison plots.

Experiment Steps

1. Define the object, the reference scan, and the accuracy metric you will use.
2. Choose one capture variable to change first, such as photo count, lighting, or camera angle.
3. Plan a repeatable capture setup so each trial keeps the rest of the pipeline as constant as possible.
4. Build a processing path that turns raw photos into a mesh you can inspect and print.
5. Decide how you will align the print or mesh to the reference scan and measure surface deviation.
6. Pre-register your comparisons so you know which result counts as a success before you start.

Common Pitfalls

• Changing room light between trials, which shifts exposure and breaks model consistency.
• Using a glossy or reflective object, which confuses feature matching and creates holes.
• Letting the turntable wobble, which changes viewpoint geometry and adds alignment error.
• Comparing scans without scaling them to the same size, which makes the error numbers meaningless.
• Cleaning the mesh too aggressively, which can erase real surface detail before you measure it.

What Makes This Competitive

A stronger version of this project would test more than one reconstruction path and compare them with the same error metric. You could pair visual quality with measured surface deviation, then separate shape accuracy from texture quality. A good entry also explains why one capture choice wins, not just which one looks best. That turns a cool demo into a real engineering study.

Project Variations

• Use a shiny historical replica instead of a matte figurine to test how reflectivity changes reconstruction quality.
• Compare phone cameras from different price tiers to see whether sensor quality changes surface-error results.
• Swap the object for a small natural specimen, like a shell or rock, and test whether irregular texture improves feature matching.

Learn More

• NASA 3D Resources: Search NASA's open 3D and visualization materials for examples of digital modeling and printable assets.
• MIT OpenCourseWare: Search for computer vision and digital fabrication course materials that explain reconstruction pipelines.
• NIH PubMed: Search for review articles on photogrammetry, 3D reconstruction, and metrology.
• USGS 3D Elevation Program: Explore how survey-grade shape data gets collected and checked.
• The 3D Printing Handbook: Find it through a library or bookstore to learn mesh prep, scaling, and print planning.
• Blender Manual: Use the free official documentation to learn mesh cleanup and export settings.