OpenCV accumulateSquare_1

Short version: accumulateSquare_1 is a duplicate of accumulateSquare_0. Identical inputs (src, dst, optional mask), identical dst = dst + src² behavior, identical single nparray output. The _1 comes from OpenCV's overload list, and for this function the overloads collapse to the same signature, so opencv-comfyui emitted two interchangeable nodes. Use whichever one you find.

What it does

It squares every pixel of src and adds the result into a floating-point accumulator, growing Σ(src²) across however many frames you feed it. Non-saturating by design - that's the whole point of the accumulate family versus add_0, which clips. Paired with accumulate_0's Σ(src), it's the raw material for frame-sequence variance and standard deviation. On its own it's a statistic you'd combine later, not an effect you'd view.

The inputs that matter

• src (NPARRAY) - the image being squared and folded in.
• dst (NPARRAY, required) - the accumulator. Must be float32, same size as src.
• mask (NPARRAY, optional) - only accumulate where the mask is nonzero.

Output: one nparray, the updated Σ(src²).

Where it fits

Realistically, in the same narrow slot as all the accumulate nodes: temporal statistics over a frame sequence, inside a workflow that already iterates frames. Stock ComfyUI doesn't loop natively, so this leans toward custom pipelines. The README's honest warning - auto-generated nodes, some not useful inside ComfyUI without extra processing - is the right frame for it. When you need variance over frames, it's exactly the right primitive; until then, it'll just sit in the menu.

Installing opencv-comfyui

ComfyUI Manager → search "opencv-comfyui", or:

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

Restart ComfyUI. OpenCV (opencv-contrib-python) is the dependency; you almost certainly have it already.

Gotchas

• Float accumulator only - uint8 dst trips dst.type() == CV_32F.
• Batch limit - Image2Nparray only handles batch_size == 1.
• Float output ≠ displayable image - don't feed the accumulator to Nparrays2Image and expect a picture back.