OpenCV convertPointsFromHomogeneous_1

convertPointsFromHomogeneous_1 is the _1 twin of a very niche projective-geometry node, and like every twin in this pack it's generated from the second overload of the same OpenCV function - the UMat version of a signature that also exists for MatLike. Both generated nodes are identical. Everything you need to know is in convertPointsFromHomogeneous_0; this page exists so the search result for the _1 suffix isn't a dead end.

Quick recap of the mechanism, because it's a two-sentence topic: the node takes src (an NPARRAY of homogeneous points), divides every point by its last coordinate, and returns the NPARRAY in Euclidean form. (x, y, z, w) → (x/w, y/w, z/w). Optional dst input is an OpenCV out-parameter - skip it. No image in, no image out, nothing to tune. It's the "undo the convertPointsToHomogeneous_0" step in a homography or projection pipeline.

Should you care? Only if your workflow does camera or projective math. For everyone else this is a nothingburger with a confusing name - the pack is auto-generated from OpenCV's type definitions (the author says the nodes are "ugly and complex to use. Expect dragons!"), and it dutifully exposes hundreds of functions whether or not they fit ComfyUI's image-centric world. This one mostly doesn't.

Install, same as the whole pack

cd ComfyUI/custom_nodes
git clone https://github.com/geroldmeisinger/opencv-comfyui
pip install opencv-python-contrib

Or ComfyUI Manager → search opencv-comfyui → restart. No models.

And the one gotcha worth remembering: this pack's nodes work on numpy NPARRAYs, not Comfy IMAGE tensors, and the outputs here are point arrays, not pictures. If you wire the output into Nparrays2Image hoping to see something, you'll get the 'NoneType' object has no attribute 'shape' error from the README. That's not a bug - it's the pack telling you this output was never an image.

Between _0 and _1: no meaningful difference. Pick whichever, wire it, move on.