OpenCV minMaxLoc_0

Need to know exactly where the brightest pixel of a mask or depth map is, and how bright it actually is? That's minMaxLoc_0 - a thin wrapper around cv2.minMaxLoc that scans an array once and hands you the minimum, the maximum, and both their pixel locations. No model, no API, no key. It's a measurement node, and it's one of the more genuinely useful ones in this pack.

Why you'd reach for it

The typical ComfyUI moment: your pipeline produced a mask, a depth map, a disparity map, or a heatmap-style array, and you have no idea what its value range actually is. Everything downstream (thresholding, normalizing, compositing) depends on that range. minMaxLoc_0 answers it in one run and also tells you where the extremes are - the brightest spot in a saliency map, the hole in a mask, the darkest shadow in a depth pass.

It pairs naturally with normalize_0: check the min/max here, then stretch the array to a known range there.

What it does

Feed it src (an NPARRAY) and it returns four things:

• float_0 - the minimum value
• float_1 - the maximum value
• literal_2 - the min's location
• literal_3 - the max's location

The mask input is optional: if you only care about a region, pass a mask and it limits the search. One caveat that catches everyone: src must be single-channel. A 3-channel RGB/BGR array throws error: (-215:Assertion failed) img.type() == CV_8UC1. Fix it with cvtColor (code=6, BGR2GRAY) before this node.

Also note the two location outputs come back as STRING literals, something like (42, 87), not as two separate floats. That's the auto-generation tax: Point is a composite type in OpenCV, and this pack hands composite types back as parseable strings. If you need the x/y as numbers, ast-parse the string or use an intermediate step.

Installing it

Part of geroldmeisinger/opencv-comfyui, a pack that auto-generates a node for ~635 top-level cv2 functions. Install via ComfyUI Manager (search "opencv-comfyui") or:

cd ComfyUI/custom_nodes
git clone https://github.com/geroldmeisinger/opencv-comfyui

then restart. The only real dependency is OpenCV (opencv-contrib-python); most installs already have it via another node pack. If you see Cannot import name 'guidedFilter' from 'cv2.ximgproc', you have conflicting OpenCV installs - that's a known problem with a documented fix in the README.

The pipeline you need first

These nodes speak NPARRAY, not Comfy's IMAGE. Chain: IMAGE → Image2Nparray → minMaxLoc_0 → ... → Nparrays2Image → IMAGE. Image2Nparray flips RGB→BGR and only takes batch_size==1 (otherwise use ImageFromBatch with length=1). Nparrays2Image flips back for viewing.

Troubleshooting

• img.type() == CV_8UC1 - you fed a 3-channel array; grayscale it first with cvtColor code 6.
• Batch error - Image2Nparray only accepts one image at a time; slice with ImageFromBatch.
• The float/literal outputs aren't images, so don't feed them into Nparrays2Image - that's where 'NoneType' object has no attribute 'shape' comes from.