Text-to-Tactile 3D Sculpture for Touch

ISEF Category: Technology Enhances the Arts

Subcategory: 3D Modeling  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A picture can help sighted people recognize an object in seconds, but touch has to do the same job in a few careful passes. That means shape, texture, and spacing all matter. Your project can turn a text prompt into a tactile sculpture and test whether people can identify it by touch. That is part art, part AI, and part human perception.

What Is It?

This project asks a simple question with a hard answer, can a computer turn words into a shape that hands can understand? You start with a text description, then use software to build a 3D model from shape primitives, which are basic building blocks like cylinders, spheres, and boxes. SDF means signed distance field, a math way to describe where a shape exists in space. Think of it like sculpting with invisible clay rules instead of a chisel.

The second part adds roughness maps. A roughness map controls where the surface feels smooth, bumpy, or ridged. That matters because touch reads contrast differently from vision. If two regions feel too similar, the object becomes hard to identify. If the texture differences are too busy, the object can turn into noise under the fingers. Your job is to balance both shape and texture so the final print carries meaning.

Why This Is a Good Topic

This is a strong science fair topic because you can test a clear question, measure real performance, and improve the design with data. You can connect it to accessibility, assistive design, and digital fabrication, which gives the project real-world value. You will also learn prompt design, 3D modeling logic, user testing, and statistics. Those skills matter far beyond one print.

Research Questions

• How does the choice of SDF shape primitive affect touch-based object recognition accuracy?
• What is the effect of roughness-map contrast on how quickly users identify a printed object by touch?
• Does combining distinct texture zones with simpler shapes improve recognition more than adding shape detail alone?
• To what extent does object size change tactile recognition for blind or low-vision users?
• Which combination of shape complexity and surface texture produces the highest recognition score?
• How does prompt wording influence the clarity of the generated tactile model?

Basic Materials

• Computer with access to 3D modeling software.
• LLM text prompt interface.
• Free or student-accessible 3D modeling software.
• Slicing software for 3D printing.
• 3D printer access, or a local makerspace printer.
• Assorted 3D print filament or resin, depending on printer type.
• Opaque labels or masking tape for participant coding.
• Blindfolds or vision-blocking goggles for sighted pilot testers.
• Stopwatch or timer.
• Data sheet or spreadsheet for recording recognition scores.
• Caliper for measuring printed dimensions.

Advanced Materials

• High-resolution 3D printer.
• Tactile evaluation set with multiple surface finishes.
• Haptic exploration stand or consistent touch-testing jig.
• Materials for post-processing and surface finishing.
• Access to blind or low-vision participant testing through approved protocols.
• Audio recorder for think-aloud comments, if permitted.
• Force sensor or pressure-sensitive input device.
• Computer with mesh-editing, prompt, and image-analysis tools.
• Scanning or metrology equipment for checking print fidelity.
• Institutional review board or school ethics approval materials, if human testing is involved.

Software & Tools

• Blender: Builds and edits the 3D mesh before printing.
• Python: Automates prompt handling, mesh generation, and scoring analysis.
• ImageJ: Measures surface texture contrast from printed or rendered test panels.
• MeshLab: Inspects mesh quality, fixes geometry issues, and checks printability.
• Google Sheets: Organizes recognition data and basic statistics.

Experiment Steps

1. Define the tactile task you want to optimize, such as identifying shapes, categories, or symbols by touch.
2. Choose the one variable you will change first, such as primitive type, roughness strength, or shape complexity.
3. Plan a repeatable way to convert text into a mesh, then keep the rest of the pipeline fixed.
4. Build a standard set of comparison objects so you can judge whether tactile changes help or hurt recognition.
5. Design a fair touch test that controls for order, time, and prior exposure.
6. Decide how you will score performance, then plan the statistics before you print the final set.

Common Pitfalls

• Making textures too subtle, which leaves the printed object looking different but feeling the same.
• Letting prompt wording change from trial to trial, which makes the generated shapes impossible to compare.
• Testing with only sighted classmates and assuming the results transfer to blind users.
• Printing models with fine details that collapse or blur during fabrication, which breaks the tactile signal.
• Changing both shape and texture at the same time, which makes it hard to tell what actually improved recognition.

What Makes This Competitive

A competitive version of this project would treat tactile design as a measurement problem, not just a cool print. You would compare multiple shape and texture strategies, use a clear statistical plan, and explain why your design helps touch perception. Strong entries often include a user-centered test with careful controls, plus an analysis of where the pipeline fails. The best projects also show an improvement over a baseline, not just a working prototype.

Project Variations

• Test whether tactile icons for classroom labels work better with geometric primitives or more organic shapes.
• Compare smooth, ridged, and dimpled roughness maps for recognizing animals, tools, or math symbols by touch.
• Evaluate whether the pipeline works better for small handheld objects or larger tabletop models.

Learn More

• Blender Manual: Official guides for modeling, mesh editing, and 3D print preparation, found through the Blender documentation site.
• MIT OpenCourseWare, Introduction to Deep Learning: Free course materials that help you understand how learned models generate structured outputs, found on MIT OpenCourseWare.
• NIH PubMed: Search for review articles on tactile perception, haptics, and accessibility design to ground your research in prior studies.
• NASA 3D Resources: Browse free model and fabrication resources to study how complex shapes are prepared for printing, found on NASA's education and 3D asset pages.
• IEEE Transactions on Haptics: Search the journal for studies on touch perception, tactile graphics, and haptic object recognition.