@smcgann/node-red-face-detection-plus

A Node-RED node that detects faces using AI.

You can use the output with 🔗 @smcgann/node-red-vectorize-plus as part of a face recognition flow. You can also use it to crop the face images from the original image and save them as files.

This node is derived from 🔗 @GOOD-I-DEER/node-red-contrib-face-detection

Key Changes

✔ Added YOLOv8s-face model.
✔ Added postinstall script to download models.
✔ Updated dependencies.
✔ Image Buffer output now includes bounding boxes (useful for annotations).
✔ Moved data and originImg into separate msg properties.
✔ Returns number of detected faces in msg.payload.
✔ Supports runtime configuration through msg.faceOptions for dynamic settings.
✔ Includes metadata in msg.faceConfig (threshold and model).
✔ More robust error handling.
✔ Better Performance.
✔ Easier integration into Node-RED flows.

Installation

Either use the Edit Menu - Manage Palette option to install, or run the following command in your Node-RED user directory - typically ~/.node-red

cd ~/.node-red/
npm install @smcgann/node-red-face-detection-plus

Restart your Node-RED instance

Input Properties

🖼️ msg.payload → Image Buffer

• Accepts PNG, JPG, GIF, WEBP, TIFF, AVIF (binary image data).

⚙️ msg.faceOptions → Object (Optional)

• Allows overriding node config settings dynamically.
• Example:

{  
  "threshold": 0.4,  
  "model": "yolov8n-face"  
}  

Node Properties

🏷 Name

• The name displayed in the Node-RED editor.

🔍 Model

Select the detection model:

• YOLOv8n (faster, good for larger faces)
• YOLOv8s (detects smaller faces, but slower)

📤 Return Value

Defines the type of data returned:

• Detected Object → data.boxes
• Image Buffer → data.faces (array of cropped face images) + data.boxes
• Image File → Saves detected faces as separate files.

📂 Absolute Path Dir

(Visible if Image File is selected)

• Defines where the cropped face images are saved.

🎯 Confidence Threshold

• Defines the minimum score required for a detected face to be considered valid (range: 0.1 - 1):

  | Threshold | Effect |
  |-----------|--------|
  | Low (0.1 - 0.4) | More faces detected, but higher risk of false positives. |
  | High (0.7 - 1.0) | Fewer faces detected, reducing false positives but possibly missing faces. |

Output

Data is output in the output format selected in the 'Return Value' property.

📌 msg.payload → Number

• Number of faces detected.

🖼️ msg.originImg → Buffer

• Input Image.

⚙️ msg.faceConfig → Object

• Returns values used for detection.

  {  
    "threshold": 0.4,  
    "model": "yolov8n-face"  
  }  

✍️ Authors

S.McGann → Modified Version.

GOOD-I-DEER (Samsung Software Academy for Youth)

• Kim Jaea
• Yi Jong Min
• Lee Deok Yong
• Lee Che Lim
• Lee Hyo Sik
• Jung Gyu Sung

📜 Copyright and license

S.McGann → Modified Version

Samsung Automation Studio Team under the GNU General Public License v3.0 license.

References

🔗 Node-RED Creating Nodes
🔗 Samsung Automation Studio Github Repository
🔗 Ultralytics YOLOv8
🔗 YOLOv8 ONNX in JavaScript
🔗 YOLOv8 ONNX in Node.js
🔗 YOLOv8 Face Model