RekogniFlow 📊 Score Display

This is the quiet one in the pack, and the only node in RekogniFlow that makes zero AWS calls. RekogniFlow_ScoreDisplay takes a raw similarity float and formats it into a human-readable string with a label: 98.1234% - EXCELLENT, 72.5% - POSSIBLE, that sort of thing. That's it. No network, no API key, no cost.

You'd think a formatter is beneath notice, but it solves a real ComfyUI annoyance. The compare nodes hand you similarity_score as a bare float, and staring at 0.9734210513 in a text box while you try to judge a result is miserable. Feed it through this node, run the string into a text display or just let it sit on the wire, and you instantly know where the likeness lands on the pack's scale.

How it works

Dead simple, and it's worth knowing the labels because they're the pack's shared vocabulary:

| Score | Label | |---|---| | 95–100% | EXCELLENT | | 80–94% | GOOD | | 50–79% | POSSIBLE | | 0–49% | NO MATCH |

The node rounds the float to your chosen decimal_places and returns a single string combining the percentage and the label. Note the forceInput on the score input - this node is designed to be wired, not typed into.

Inputs and output

• similarity_score (FLOAT) - plug in the similarity_score output from RekogniFlow_FaceCompare (or best_score from the batch compare).
• decimal_places (default 4, 0–10) - precision. Drop it to 1 or 2 if you want 98.1% - EXCELLENT instead of the full precision.
• Output: formatted_score (STRING) - the labeled string. Wire it to any STRING/text display node.

Installing

Same pack as the whole RekogniFlow suite, so if you've installed any of the others you already have this one. Otherwise: ComfyUI Manager search comfyui-rekogniflow, or clone it in:

cd ComfyUI/custom_nodes
git clone https://github.com/Aiconomist/comfyui-rekogniflow
cd comfyui-rekogniflow
pip install -r requirements.txt

Restart ComfyUI and it appears under image/face/aws (display name "RekogniFlow 📊 Score Display").

Gotchas

Almost none - this is a pure local string formatter, so the usual AWS headaches don't apply. The one thing to keep in mind: the labels are hardcoded to this pack's thresholds, so if you consider 70% "close enough," the POSSIBLE label will keep telling you otherwise. It's a label, not a verdict; use it as a quick read, not a final say. And if you only need the raw number for a numeric comparison downstream, skip this node and wire similarity_score straight through - it's purely for eyeballs.