Fabric Shrinkage Measurement Software

I am building a small, stand-alone system that lets a quality-control operator lay a piece of fabric under a camera, press one button, and instantly see the percentage of length-wise and width-wise shrinkage after washing.

Here is what I need you to put together:

• Vision workflow
A single industrial or high-resolution USB camera points at the specimen. Your software must automatically detect the four reference marks I place on the cloth, both in the “before wash” and “after wash” images, and then calculate the change in X and Y with an accuracy of ±0.5 mm. The algorithm should work reliably on both knitted and woven textiles, even when the background and yarn colours vary. OpenCV (Python) contour / blob / template methods are fine as long as they meet the tolerance.

• Desktop application on Windows
A lightweight Windows executable is all I need. You are free to use PyQt, Tkinter or another GUI wrapper, as long as the installation is simple and offline-friendly.

• Clear, text-based result display
Once the second image is captured the app should immediately show a clean text summary such as
“Length shrinkage: 3.2 % Width shrinkage: 2.7 %”.
No dashboards or plots are necessary, but the code should leave room for future extensions.

• Automation hooks
Please expose a command-line flag or minimal API so the measurement routine can be triggered from an external PLC later on.

Acceptance will be based on:
1. Repeatable detection of the four marks on at least 30 consecutive samples with ≤ 0.5 mm error, verified by a calibrated ruler.
2. Executable installer and full Python source, commented and ready to re-build.
3. A brief user manual covering camera setup, calibration procedure, and typical runtime (target < 5 s per image pair).

If you have solid experience in Python, OpenCV, computer-vision calibration, and basic desktop GUI work, this should be a concise yet interesting project. I’m ready to start as soon as you can outline your chosen detection approach and estimated timeline.