ComfyUI Node
arkennemasis Calibrate (derive tolerances from a real set)
Run every check across an EXISTING set of images and derive thresholds from measurement instead of guesswork. No generation, no spend. With operator labels it picks thresholds that separate accepted from rejected and reports the true current pass rate — which is a number worth having before building anything.
arkennemasis Calibrate (derive tolerances from a real set)
- plate
- report
- tolerances_json
- frame_tolerance
- colour_tolerance_de
- identity_tolerance
◄image_dir►
◄plate_lock_json►
◄labels_csv►
◄targets_csv►
◄percentile95.0►
◄limit0►
◄write_to►
Categoryarkennemasis/Variation
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| image_dir | STRING | Folder of already-produced images for this product. | |
| plate | IMAGE | The locked base plate. | |
| plate_lock_jsonopt | STRING | — | |
| labels_csvopt | STRING | Optional CSV of 'filename,label' where label is accept or reject. With it, thresholds are chosen to separate the two populations; without it the report is descriptive only. | |
| targets_csvopt | STRING | Optional CSV of 'filename,region,hex' so the colour check has a per-image target. | |
| percentileopt | FLOAT | 95.050–100 | With no labels, thresholds are set at this percentile of the observed distribution. |
| limitopt | INT | 00–10000 | Measure at most this many images. 0 = all. |
| write_toopt | STRING | Optional folder for calibration.json and the per-image measurements CSV. |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| report | STRING | — |
| tolerances_json | STRING | — |
| frame_tolerance | FLOAT | — |
| colour_tolerance_de | FLOAT | — |
| identity_tolerance | FLOAT | — |