OpenCV calcCovarMatrix_1

OpenCV calcCovarMatrix_1 is the second overload of cv2.calcCovarMatrix, and it is functionally identical to calcCovarMatrix_0. opencv-comfyui generated a node for every overload it found in OpenCV's type stubs, and this function has two. There is nothing to choose between them - pick one, and treat the other as pack background noise.

For the rest of the story, read the twin article; the short version: this is a statistics function, not an image filter. It takes a set of samples and returns the covariance matrix (how every dimension co-varies with every other) plus the mean - the raw material for PCA, whitening, and Mahalanobis-style analysis. It belongs to the class of OpenCV functions the pack's own README flags as "not every function is useful within ComfyUI without further processing." For 99% of ComfyUI users, this is the node you'll never wire - and that's the correct call.

Inputs and outputs

• samples - NPARRAY: your data, rows vs. columns meaning samples vs. features depending on the flags.
• mean - NPARRAY: an OpenCV out-parameter that the generator surfaced as a required input. It gets overwritten with the computed mean, but you still have to connect something to satisfy the socket - one of the pack's known auto-gen awkwardnesses. Feed any same-shape array.
• flags - INT: cv2.COVAR_* mode (1 = CV_COVAR_NORMAL, 0 = CV_COVAR_SCRAMBLED, etc.).
• ctype - INT: output array type (5 = CV_32F, 6 = CV_64F).
• covar (optional) - the other out-parameter; leave unwired.

Outputs: nparray_0 (the covariance matrix) and nparray_1 (the mean) - square float matrices and a vector, not images. Sending them into Nparrays2Image produces exactly the 'NoneType' object has no attribute 'shape' error the README warns about.

Install

ComfyUI Manager → opencv-comfyui, or clone https://github.com/geroldmeisinger/opencv-comfyui into ComfyUI/custom_nodes, restart. Needs opencv-contrib-python (near-certain already-have). Batch size 1 and the NPARRAY color/depth conventions apply as everywhere in this pack, though they matter little for pure matrix math.

Bottom line: if PCA-style statistics inside a ComfyUI graph is your thing, _0 or _1, it doesn't matter which. If it isn't, you've now got one fewer mystery when you scroll past it in the node list.