From one photograph to a relief: why the depth map’s bit depth shows up in the part
A measurable comparison of a genuine 16-bit depth map and an 8-bit file of the same height field, using the public lion-medallion sample from Depth-Map.com.
The same relief, two bit depths
Shaded previews exaggerate height so the steps are visible. These two illustrations are 8-bit RGB pictures of the height field, not the manufacturing files themselves.
What we measured (15 November 2026)
| File | PNG header | Unique gray values | Min / max |
|---|---|---|---|
| Genuine 16-bit manufacturing PNG | 16-bit grayscale | 19,662 | 0 / 19,661 |
Same field saved as 8-bit (value // 257) | 8-bit grayscale | 77 | 0 / 76 |
8-bit written back into a 16-bit container (value * 257) | 16-bit grayscale | 77 (all multiples of 257) | 0 / 19,532 |
Depth-Map.com writes depth on a 10 mm scale: 65,535 maps to 10 mm. A 3 mm relief therefore uses less than a third of the range, so the brightest pixel is 19,661, not 65,535. That is expected. A fake 16-bit file is the one whose header says 16-bit but whose unique-value count is 256 or fewer.
How to check a file
from PIL import Image
im = Image.open("lion-depth-16bit.png")
print(im.mode) # I;16
print(len(set(im.getdata()))) # 19662
print(min(im.getdata()), max(im.getdata())) # 0 19661
Or use ImageMagick: identify -verbose your-depth-map.png and read the gray channel min/max. The source files, hashes, and a small verify.py script are in the companion dataset.
What to do with the files
- Send the 16-bit PNG to LightBurn in 3D Sliced mode.
- Import the closed STL into a slicer or a CAM tool. The site’s published sample was exported at 3 mm relief on a 1.5 mm base, with invert on.
- If you need a mechanical part with exact dimensions, model it in CAD. A single photograph is a height field, not a dimensionally accurate solid.
Files and measurements: the companion dataset depthmap/16bit-vs-8bit-depth-maps (published with this Space).
Longer write-up: 16-bit vs 8-bit guide on Depth-Map.com.
Generate your own 16-bit PNG and closed STL from a photo, including this sample, at depth-map.com. Three photos a day do not require an account.