Fine Crack Tracing Toolkit
Turns cracks already found in an image into ordered, measurable paths, with overlays and metrics.
My partMaintainer
In short
An installable Python package and CLI for turning pre-segmented crack masks into ordered lines, optional splines, visual overlays, and evaluation metrics.
- What it is
- A Python package and command-line interface that turn pre-segmented crack masks into ordered paths, smoothed curves, overlays, and metric exports.
- Why it matters
- A segmentation mask names the pixels but leaves inspection engineers without an ordered trace, a comparable reference, or an evaluation record.
- My part
- I own and maintain the package, its command-line interface, and its documentation; the algorithms it builds on are credited to their sources.
Step by step
- 1
Ingest
The toolkit consumes raw frames and pre-segmented masks. It begins after segmentation and does not attempt to detect cracks itself.
- 2
Order
Candidate points can be ordered with a classic approach, a minimum-spanning-tree path, or a greedy alternative, chosen for the geometry at hand.
- 3
Smooth
Optional spline fitting produces smoother traces without erasing meaningful bends. ORB or Shi-Tomasi features support corner detection where local structure matters.
- 4
Evaluate
Outputs can be evaluated with precision, recall, F1, IoU, mean and maximum distance, and RMS error, and exported to CSV and JSON.
For engineersHow it is built, its limits, and its keywords.
How it works
The toolkit accepts frames and masks, extracts candidate crack points, orders them with a choice of classic, minimum-spanning-tree, or greedy strategies, optionally fits smoother curves, generates overlays, and exports metrics as CSV and JSON.
Limits
- The toolkit begins after segmentation and is not a crack detector.
- Validation at dataset scale, across crack types, widths, and imaging conditions, is the next step.
- The metrics keep their standard definitions; the package does not redefine them.
Keywords
- crack tracing
- segmentation post-processing
- infrastructure inspection
- ordered paths
- evaluation metrics