Skip to content

synthetic-ndt-spike.md

Latest commit

 

History

History
65 lines (47 loc) · 2.29 KB

synthetic-ndt-spike.md

File metadata and controls

65 lines (47 loc) · 2.29 KB
Raw

Synthetic radiograph spike

DataLab-Web ships public/demos/ndt.h5 as a deterministic synthetic NDT spike. It is algorithm-development data, not an industrial radiograph, a validated detection workflow, or evidence of compliance with an inspection standard.

Scope

The generator in scripts/ndt_synthetic.py composes:

  • a smooth background representing thickness variation;
  • seeded detector noise and low-amplitude texture;
  • vertical and horizontal detector artifacts plus isolated bad pixels;
  • compact and linear indications;
  • varied indication sizes, orientations, and contrasts;
  • one declared pair of overlapping compact indications.

The output is deterministic for a given image size and seed. The default 512 by 512 image is stored as unsigned 16-bit data.

Ground truth

The workspace stores JSON ground truth in the image metadata key spike.ndt.synthetic_truth. Each indication has a stable identifier and records:

  • class: compact or linear;
  • position: center coordinates in pixels;
  • geometry: ellipse radii and orientation;
  • contrast: signed intensity difference used by the generator.

The payload also records the generator seed, image size, declared overlaps, and artifact coordinates. spike.ndt.disclaimer marks the image as synthetic algorithm-development data. The Python test suite opens the shipped HDF5 file through the production workspace loader and verifies that these fields survive the round trip.

Permitted use

The spike may support tests of geometry, deterministic algorithm behavior, detection rate, false-positive count, and measurement error against its own synthetic truth. This repository does not currently include or validate an NDT detection algorithm against those metrics.

The spike does not validate:

  • radiographic realism or acquisition physics;
  • performance on industrial data;
  • an inspection protocol or image-quality threshold;
  • automatic accept/reject decisions;
  • compliance with an NDT standard.

Product development remains gated on an NDT subject-matter expert and an exploitable annotated real-world corpus.

Regeneration

Run the workspace generator in an environment containing Sigima and guidata:

python scripts/gen_demo_workspaces.py

The script also regenerates the spectroscopy and photonics demo workspaces.