INDUSTRY FIELD GUIDE
Nanotechnology microscopy scale and measurement evidence for AI
A visually similar particle can have a different measured size when images use different magnification or preparation.
NAICS 541713 · Research and Development in Nanotechnology
A visually similar particle can have a different measured size when images use different magnification or preparation. NIST publishes methods for measuring nanoparticle size with transmission electron microscopy. A laboratory can create a low-risk interpretation test using independently generated images and known geometry. NIST nanoparticle microscopy measurement research
Generate two synthetic images with known particle outlines and different scale bars. Crop one image so its scale is missing. Ask which comparisons have measurement support and where the estimate should be withheld. Include a separate condition label when particles overlap in a view.
Use generated shapes and original metadata. Exclude sponsor data, unpublished formulations, hazardous sample preparation and proprietary instrument settings. Keep validation with qualified personnel.
Describe a synthetic microscopy evidence sample
Original Fiund exercise
The same pixel width can represent different dimensions under different scale bars
Independently drawn circles replace microscopy images. Practice units are arbitrary, with known flat geometry and no material, formulation or preparation procedure.
| Record | Source | What it records |
|---|---|---|
| NM1 | Synthetic image A | Circle spans 40 pixels. Scale bar 20 pixels=10 practice units; outline unobstructed. |
| NM2 | Synthetic image B | Circle spans 40 pixels. Scale bar 20 pixels=5 practice units; outline unobstructed. |
| NM3 | Cropped image C | Circle-like outline spans 40 pixels, scale bar absent and another shape overlaps one edge. No parent-image link supplied. |
Question for the model
What dimensions follow for A and B, and can C be assigned the same size from pixel width or resemblance alone?
Show the reference answer and checks
NM1 supports A's 40-pixel span as 20 practice units. NM2 supports B's same pixel span as ten units because its scale differs. NM3 lacks the scale and a complete unobstructed outline, so neither size can be transferred from resemblance or pixel count. Preserve scale provenance and overlap flags rather than treating magnification-independent dimensions as directly visible in every image.
- Comparing pixel widths as physical dimensions without each image's scale.
- Estimating a full outline from an overlapping shape or an unlinked cropped image.
Fields to preserve
- Image identity and scale bar
- Defines the pixel-to-unit conversion.
- Measured span and outline visibility
- Makes the synthetic geometry check reproducible.
- Crop parent and overlap condition
- Preserves missing source or shape evidence.
Test another case
Link C to a known parent but crop through the scale and particle boundary. Require scale lineage and measurable-outline completeness to be checked separately.
Use original generated geometry and metadata. Exclude sponsor data, formulations, hazardous preparation and proprietary instrument settings.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.