OpenCV computeCorrespondEpilines_1
ComputeCorrespondEpilines_0, duplicated
- points
- F
- lines
- nparray
computeCorrespondEpilines_1 is the duplicate of computeCorrespondEpilines_0. Same cv2.computeCorrespondEpilines call, same inputs, same output, byte-for-byte identical implementation aside from the name. The pack generates a node per overload in OpenCV's type stubs, the function declared two signatures, and here you are with two identical nodes. Use the _0 one; this is a spare.
The short version
Stereo-geometry helper: given points in one image of a stereo pair and the 3×3 fundamental matrix F, compute the epipolar line in the other image where each match must lie.
- points (NPARRAY) -
(N, 1, 2)or(N, 2)of(x, y)pairs. - whichImage (INT) -
1or2, saying which image the points live in. Getting this backwards gives silently wrong lines, so check it. - F (NPARRAY) - the fundamental matrix, typically from
findFundamentalMat(also in this pack). - lines (NPARRAY, optional) - out-parameter; skip it.
- Output:
nparray-(N, 1, 3)lines(a, b, c)fora*x + b*y + c = 0.
This is real 3D-vision math - structure-from-motion, matching outlier rejection, rectification. In a pure image-gen workflow it's niche, and it assumes you already have point correspondences.
Install
Same as every node in this pack:
cd ComfyUI/custom_nodes
git clone https://github.com/geroldmeisinger/opencv-comfyui
or ComfyUI Manager → "opencv-comfyui". Needs opencv-python-contrib. Watch for the Cannot import name 'guidedFilter' from 'cv2.ximgproc' conflict if multiple OpenCV installs collide. And one honest note: if you landed here because you saw both _0 and _1 in your node search and wondered which to use - either one. They're the same node.
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| points | NPARRAY | — | |
| whichImage | INT | — | |
| F | NPARRAY | — | |
| linesopt | NPARRAY | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| nparray | NPARRAY | — |