TAMP-OS: Tactile Accessible Microscopy Printing (Open-Source)

Natalia Gonzalez-Vazquez, Robert Faulkner, Victoria Gamez, Karly E. Cohen, Gunther Richter, Abigale Stangl, Andrew K. Schulz

TAMP-OS is an open-source extension and update of the original TAMP (Tactile Accessible Microscopy Printing) workflow for converting microscopy images into 3D-printable tactile lithographs. The original TAMP workflow was used by the co-authors to create 3D-printable lithographs of scanning electron microscopy images with additional details found in the supplemental materials of the Science publication.

After the publication, we received requests to make the workflow more streamlined and open-source, which is this extension page. This updated workflow has been submitted for publication and is currently under review, entitled "TAMP-OS: Tactile accessible microscopy printing workflow for open-source creation of 3D printable lithographs", by Natalia Gonzalez-Vazquez, Robert Faulkner, Victoria Gamez, Karly E. Cohen, Gunther Richter, Abigale Stang, and Andrew K. Schulz. Below there are easy access buttons to the original publication, preprint for TAMP-OS, and Edmond repository of the microscopy data used in this publication.

---

What is TAMP-OS

Built on TAMP — the original Tactile Accessible Microscopy Printing workflow (Faulkner et al., Science 2026), which converts microscopy images into 3D-printed tactile lithographs using Bambu Lab hardware. The original workflow, figures, and README are preserved in tamp_original/.

TAMP makes scientific imagery accessible to blind and visually impaired people by converting microscopy images into 3D-printed tactile lithographs. The original workflow depended on Bambu Lab hardware and the proprietary Bambu Maker Lab slicer.

TAMP-OS replaces that proprietary dependency with a fully open-source pipeline that works with any FDM printer, gives full scripting control, and requires no hardware lock-in.

Microscopy Image → Height Map → STL → G-code → Any open-source printer → Tactile Lithograph

---

Goals

• Make tactile science communication accessible to any lab, regardless of budget or hardware
• Provide a fully scriptable, automatable pipeline
• Be approachable for researchers with no programming experience
• Support multiple output formats (STL, 3MF, GLB) and printers

---

Fabrication method

This pipeline produces discrete extruded topology — height maps converted to layered FDM (fused deposition modeling) prints. The nozzle diameter and layer height define the minimum feature size that can be physically reproduced.

This is distinct from continuous microtopography fabrication methods (e.g. photopolymer jetting, SLA, or specialized tactile printing systems), which can achieve finer surface gradients. If you are using a non-FDM printer, the nozzle-based preset calculations will not apply — use the Full customization panel and the sweep tool to determine your own optimal parameters.

---

Why open-source?

The Bambu Lab ecosystem has three limitations for research use:

• Proprietary slicer — no scripting or automation support
• No API control — cannot integrate into larger pipelines
• Hardware lock-in — costly and not community-repairable

This pipeline uses only open-source tools and works with any FDM printer.

---

Tools

Main tool

File	Format	Description
tamp_batch_gui_v2.ipynb	Jupyter notebook	Start here. Batch GUI with Low/Medium/High quality presets calculated from your printer's nozzle and layer height. Supports .STL, .3MF, .GLB output.

Note: Requires Jupyter Lab or Jupyter Notebook - not VS Code notebook viewer.

Supporting tools

File	Format	Description
tamp_resolution_compare.ipynb	Jupyter notebook	Quality checking tool. Run one image at multiple resolution, relief height, or blur values to find the best settings before a full batch run
pyvista_image-generator.ipynb	Jupyter notebook	3D visualization tool. Renders every STL in your comparison folder as a 3D screenshot (full view + zoom), useful for figures and presentations

---

Getting Started for non Python users

Follow these steps if you have never used Python before.

1. Install Python

Download Python 3.10 or newer from: https://www.python.org/downloads/

⚠ On Windows, check "Add Python to PATH" during installation.

Verify it worked — open a terminal and type:

python --version

You should see something like Python 3.11.2.

2. Install dependencies

Open a terminal, go into the repo folder, and run:

pip install -r requirements.txt

You only need to do this once.

3. Install Jupyter Lab (if you don't have it)

pip install jupyterlab
jupyter lab

4. Open the main tool

In Jupyter Lab, open tamp_batch_gui_v2.ipynb and run all cells. The GUI window will appear.

---

The Main Batch GUI (v2)

The v2 GUI is designed around a simple print workflow: add images, choose a printer profile, choose Low / Medium / High quality, and generate print-ready files. The profile represents the machine capacity, while Full customization keeps the detailed printer and mesh settings available when needed.

How to use it:

1. Click Add images to select one or more microscopy images
2. Choose a printer profile that roughly matches the printer's capacity
3. Select a quality preset:

Preset	What it means
Low	1 pixel = 2 nozzle width. Smooth, small file. Good for quick checks.
Medium	1 pixel = 1 nozzle width. Matches what your printer can actually reproduce. Recommended starting point.
High	1 pixel = 0.5 nozzle width. Finer than the nozzle — captures more texture, larger file.

Each preset card includes an approximate binary STL size. The estimate is based on the generated mesh triangle count, so it changes with resolution and image aspect ratio.

By default, one quality setting applies to the whole batch. For mixed image sets, check Choose different quality for different images to reveal a per-image table with suggested quality and a Low / Medium / High override.

4. Choose the output folder and output format
5. Click Generate print files

The default output is .STL. Open Full customization only when you need to change print size, nozzle diameter, layer height, relief height, blur, resolution, base thickness, flip, or invert relief.

---

Quality Checking Before a Full Batch Run

Before processing many images, use tamp_resolution_compare.ipynb to find the best settings for your specific image and printer.

It runs one image at multiple values of one parameter (resolution, relief height, or blur) and saves clearly named files — e.g. elephant_resolution_128.stl, elephant_resolution_256.stl — so you can compare them side by side in MeshLab or PrusaSlicer.

Recommended workflow:

1. Run tamp_resolution_compare.ipynb on one representative image
2. Open the output STL files in MeshLab to compare, or use pyvista_image-generator.ipynb to render them all as 3D screenshots automatically
3. Pick the settings that look best
4. Use those settings in tamp_batch_gui_v2.ipynb for the full batch

The figure below shows an example comparison matrix — each row sweeps one parameter while holding the others fixed. This makes it easy to isolate the effect of blur, relief height, and resolution independently.

left column: blur sweep (resolution=256, relief=3.0 mm). Middle column: relief height sweep (resolution=256, blur=1.2). Right row: resolution sweep (relief=3.0 mm, blur=1.2).

---

3D Visualization of Comparison Results

After running tamp_resolution_compare.ipynb, use pyvista_image-generator.ipynb to render all your output STLs as 3D images without opening MeshLab manually.

It reads every .stl file in your comparison output folder and saves two renders per file:

• Full view — the whole lithograph from an isometric angle
• Zoom view — a close-up of the center to show surface detail

Both dark and light themes are available. The output PNGs can be used directly in figures and presentations.

How to use it:

1. Run tamp_resolution_compare.ipynb first to generate the STL files
2. Open pyvista_image-generator.ipynb
3. Set STL_FOLDER to your comparison output folder
4. Choose THEME = "dark" or THEME = "light"
5. Run the cell — renders are saved to individual_renders/ and zoom_renders/

⚠ Install dependencies with: pip install pyvista trame

---

Choosing Parameters Based on Your Printer

The Low / Medium / High presets in the v2 GUI are calculated from your printer's nozzle diameter and layer height. The table below shows what those presets translate to for the most common nozzle sizes, assuming a 100 mm wide print.

The right choice depends on your machine, material, and time available — but the sweep tool makes it easy to verify visually before committing to a full batch.

The GUI now gives a first-pass recommendation from the image itself. Large images with strong edge detail often justify High resolution, smooth or low-pixel-count images may be fine at Low, and clipped or low-contrast images should be checked with a parameter sweep because flat tones can become flat tactile plateaus.

Nozzle	Preset	Resolution	Blur	1 px on print	Notes
0.4 mm	Low	128 px	2.0	0.78 mm	Fast print, smooth feel, small file
0.4 mm	Medium	256 px	1.2	0.39 mm	Matches nozzle — recommended starting point
0.4 mm	High	512 px	0.8	0.20 mm	Finer than nozzle, captures more texture
0.6 mm	Low	64 px	2.0	1.56 mm	Very smooth, very fast
0.6 mm	Medium	128 px	1.2	0.78 mm	Matches nozzle
0.6 mm	High	256 px	0.8	0.39 mm	Finer than nozzle
0.2 mm	Low	256 px	2.0	0.39 mm	Matches a 0.4 mm Medium
0.2 mm	Medium	512 px	1.2	0.20 mm	High detail, large file
0.2 mm	High	512 px	0.8	0.20 mm	Capped at 512 px (GitHub file size limit)

💡 These are starting points — the optimal settings also depend on the image itself. A high-contrast SEM image (sharp edges, lots of noise) may need more blur even at Medium resolution. A smooth fluorescence image may look great at Low. Use tamp_resolution_compare.ipynb to check your specific image before running a full batch.

💡 If you want full control, the Full customization panel in the v2 GUI lets you override resolution, blur, and base thickness independently of any preset. You can also work directly with the STL and adjust slicing settings in PrusaSlicer for further control.

---

Workflow Steps

1. Prepare your image in ImageJ/Fiji

• Crop out any scale bars, metadata bars, or annotations at the bottom of the image — these will appear as raised features in the lithograph
• Export as .PNG or .JPG

💡 The pipeline converts your image to grayscale automatically. If you want finer control over which features become raised vs recessed (e.g. isolating a specific channel or adjusting contrast manually), convert to grayscale in ImageJ/Fiji before exporting — but this is optional, not required.

💡 For images with very fine detail, you can apply a Gaussian blur (radius = 1.0 pixel) in ImageJ/Fiji before exporting, or simply adjust the blur parameter in the GUI.

---

2. Check your image dimensions and set print size

• Your physical print size must match the image aspect ratio — the print does NOT have to be square
• Leave print height as auto and the tool calculates it for you
• If you set it manually and it doesn't match, the script will warn you:

[WARNING] Aspect ratio mismatch!
    Height map is 512×384 (ratio 1.333)
    Print size is 100×100 mm (ratio 1.000)
    The lithograph will appear stretched. Consider size_y=75.0

Image pixels	Correct print size
1024 × 1024 (square)	size_x=100, size_y=100
1024 × 768 (4:3)	size_x=100, size_y=75
1920 × 1080 (16:9)	size_x=100, size_y=56.3

---

3. Generate the STL

Open tamp_batch_gui_v2.ipynb in Jupyter Lab, add your images, choose your quality preset, and click Generate STLs.

---

4. Slice and print

Manual (any slicer):

• Load the .stl into PrusaSlicer, OrcaSlicer, or Cura
• Recommended FDM settings:
  • Layer height: 0.12 mm
  • Nozzle: 0.4 mm
  • Infill: 15% (gyroid)
  • Supports: none needed (flat base)

---

Recommended Printers

Printer	Firmware	Notes
Prusa MK4 / XL	Marlin	Most plug-and-play, good for labs
Voron 2.4	Klipper	Best for full automation via Moonraker API
Any Marlin/Klipper printer	—	Works with PrusaSlicer profiles

---

Troubleshooting

python: command not found / pip: command not found → Python is not installed or not in PATH. Re-install from https://www.python.org/downloads/ and check "Add Python to PATH" on Windows.

ModuleNotFoundError: No module named 'numpy' (or pillow, scipy, stl) → Run pip install -r requirements.txt from the repo root.

The STL looks stretched or squished → Print height doesn't match the image aspect ratio. Leave size_y as auto or check the warning message for the suggested value.

The STL is mirrored → Make sure "Flip vertically" is checked in the GUI, or that --no-flip is not set in the CLI.

The relief is too subtle or too extreme → Adjust relief height. Start with 3.0 mm. Use tamp_resolution_compare.ipynb to test values before a full batch run.

The GUI window doesn't appear (Jupyter) → Use Jupyter Lab or Jupyter Notebook. Does not work in VS Code's notebook viewer.

PrusaSlicer not found → Pass the full path with --prusaslicer. See examples in Step 4 above.

STL file is too large for GitHub (>25 MB) → Use the Low preset, or reduce --resolution to 256 in the CLI.

---

Example Output

Input: SEM_5um_raw.png — FePt spherical particles, 5 µm scale

---

Files

File	Description
tamp_batch_gui_v2.ipynb	Main tool — batch GUI with quality presets
tamp_resolution_compare.ipynb	Supporting — quality checking before batch runs
pyvista_image-generator.ipynb	Supporting — 3D rendering of comparison STLs for figures
requirements.txt	Python dependencies

---

Pipeline Details

All notebooks are self-contained — no external scripts needed. The pipeline runs these steps internally:

Step	Function	Description
1	image_to_heightmap	Grayscale → contrast stretch → Gaussian blur → flip → [0,1] float array
2	heightmap_to_stl	Builds watertight solid mesh: top relief + flat base + 4 side walls
3	Format conversion	Converts STL to .3MF or .GLB if selected

---

Citation

If you use TAMP-OS in your work, please cite:

@article{faulkner_tamp_2026,
  title   = {A low-data, low-cost, and open-source workflow for 3D printing lithographs
             for digital accessibility of microscopy images},
  author  = {Faulkner, Robert and Gonzalez-Vazquez, Natalia and Gamez, Victoria and
             Cohen, Karly E. and Richter, Gunther and Stangl, Abigale and Schulz, Andrew K.},
  journal = {Science},
  year    = {2026},
  doi     = {10.1126/science.adx8981},
}

---

Acknowledgements

The authors are grateful for the support of the Tactile Media Alliance. A.K.S. acknowledges support from the Alexander von Humboldt Foundation. G.R. and A.K.S. acknowledge support from the Max Planck Society.

---

Contact

Authored and maintained by:

• Natalia Gonzalez-Vazquez — https://github.com/nagova
• Andrew K. Schulz — https://github.com/Aschulz94

For questions or issues, open a ticket on GitHub.

If you find this repository useful, consider giving it a ⭐

---

License

This project is licensed under the GNU GPL version 3 — see the LICENSE file for details.

Copyright

© 2026, Max Planck Society