N-10 Face Detect
JHPixelProFaceDetect wraps MediaPipe FaceLandmarker and returns a lightweight 5-point landmark set plus bounding-box metadata. It is the pack's entry node for portrait alignment: the JSON landmarks output can be pasted directly into the N-06 facial aligner workflow, and the bounding-box output is useful for ranking or filtering faces in multi-subject scenes.
Schema
| Name | Kind | Type / default | Description |
|---|---|---|---|
image |
Input | IMAGE |
Source image for face detection. |
mode |
Widget | COMBO, default single_largest |
Returns one largest face or the top-K faces ranked by bounding-box area. |
max_faces |
Widget | INT, default 1 |
Maximum faces to return when mode = multi_top_k. |
confidence_threshold |
Widget | FLOAT, default 0.5 |
MediaPipe minimum detection confidence gate. |
landmarks_json |
Output | STRING |
JSON list of 5-point landmarks in [L-eye, R-eye, nose-tip, L-mouth, R-mouth] order. |
bbox_json |
Output | STRING |
JSON list of bounding boxes and metadata per detected face. |
face_count |
Output | INT |
Count of faces returned after filtering. |
Workflow preview
Workflow JSON: workflows/S-10-face-detect.json