ComfyUI Node
CV Save Classifier
Writes a trained classifier to ComfyUI/models/cvml as a gzip-compressed '.yml.gz' file, so a model trained once ('CV Train Classifier' model_yaml output) can be reloaded in a LATER workflow by 'CV Predict Classifier'. The content is the standard OpenCV StatModel YAML (KNN/SVM models embed their training data - gzip shrinks them ~10x); to use it outside ComfyUI, gunzip first, then cv2.ml.SVM_load & co. read it. An existing file of the same name is overwritten (deterministic re-runs). An EMPTY model_yaml (a trainer that reported success=false) writes nothing and flags saved=false instead of raising. This is an output node: it runs even with nothing wired downstream.
CV Save Classifier
- filename
- saved
◄model_yaml—►
◄filenameclassifier►
Categoryimage/CV/ml
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| model_yaml | STRING | The model_yaml output of 'CV Train Classifier' (loadable OpenCV YAML text). | |
| filename | STRING | classifier | File name inside models/cvml; '.yml.gz' is appended if missing, path components are stripped. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| filename | STRING | Name of the file written inside models/cvml - the same string the 'model' combo of 'CV Predict Classifier' lists after a page reload ('' when nothing was saved). The folder it lives in is deliberately NOT part of it: an absolute path leaving the node would end up in shared workflow JSON and pasted previews, carrying the drive layout, install location and user name with it. |
| saved | BOOLEAN | False when model_yaml was empty (the trainer failed upstream) - nothing was written. |