Line Length Calculator

If you've ever needed to know "how much total line is in this diagram" - for estimating laser cut time, material usage, or just QC on a drawing - you know that counting black pixels lies to you. A 2px-thick line counts double a 1px line even though they're the same length. Line Length Calculator fixes that properly: it thins every line to a single pixel, then measures the skeleton with a diagonal-aware ruler so a diagonal step counts as √2 pixels instead of a flat 1.0. The result is a length in pixels that's actually proportional to distance, not to ink.

How it works

The pipeline: composite any alpha to a white background, convert to grayscale, binarize at threshold (default 128 - pixels darker than this count as line), then skeletonize with scikit-image's skeletonize so all lines are 1px wide. Measurement then walks the skeleton and, for each pixel, checks its eight neighbors: orthogonal neighbors contribute 1.0 per link, diagonal neighbors contribute √2. Each link is counted from both ends and halved, so you get a true path length rather than a pixel count.

The skeleton_pixel_count output is that raw count - the "all pixels" number - and total_length_px is the corrected one. The gap between them is exactly the diagonal correction, and seeing both is what makes the number defensible: a 45° line comes out at ~1.41× its pixel count, which is correct.

invert (default off) flips the logic for white-on-dark images.

What comes out

• skeleton_overlay (IMAGE) - the original image with the measured skeleton drawn in red. This is your QA pass: if the red lines don't match the drawing, the number is garbage, and you'll see it immediately.
• total_length_px (FLOAT) - the headline number.
• skeleton_pixel_count (INT) - raw skeleton pixels.
• report (STRING) - a formatted text dump: image size, threshold, black-pixel count pre-skeleton, skeleton count, and final length. Wire it to a Show Text node and you've got an audit trail for whatever you're estimating.

Where it fits

Load Image → Extract Black → LineLengthCalculator → Show Text. The red overlay is the whole reason to trust the float. And if you're doing physical estimates, the numbers give you pixels-per-unit once you know one real dimension.

Install

Part of ComfyUI-HappNodeSet (mikemojen). Manager: search HappNodeSet. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/mikemojen/ComfyUI-HappNodeSet.git
pip install -r ComfyUI-HappNodeSet/requirements.txt

Restart ComfyUI. Standard pack deps - scikit-image does the skeletonization, OpenCV handles the binarization, nothing exotic.

Common issues

The threshold is the knob that matters: too high and faint lines don't make the binary mask; too low and paper texture becomes "line." The skeletonization assumption is that lines are thin - a thick solid shape will skeletonize to a single centerline, which is usually what you want, but a filled circle will measure as a diameter-ish line rather than a circumference, so don't feed it solid shapes and expect contour lengths. And with curved lines, the straight-step approximation underestimates slightly - fine for estimates, wrong if you need engineering-grade precision on tight curves. Check the red overlay once, trust it after.