A custom node from node-red to detect face masks from an input image.

🚧 Requirements

• On windows, You need the Visual Studio Build Tools because node-gyp requires to run correctly, You can read more about it here.

You can install the Build tools (2017) 👉 Official Build Tools 2017.

You need check and install these packages:

An alternative to installing the build tools is using the next command suggested here running a Powershell terminal in administrator mode.

npm install --global --production windows-build-tools

🔧 Install

This node requires @microsoft/customvision-tfjs-node and @tensorflow/tfjs-node as peer dependencies, you need installing them manually.

Go to .node red folder (usually located at C:\Users\{YOUR_USER}\.node-red) and execute:

npm install @microsoft/customvision-tfjs-node

npm install @tensorflow/tfjs-node

and then, install the node using:

npm install -S node-red-contrib-facemask-detector

or search it on Manage Pallete from node-red.

🔧 Usage

Pass an image (array buffer) to the input of node, and receive:

• The prediction model, that includes: bbox array, prediction array, and label array (0 = mask object)
• Original image processed.
• Boolean value (detected) any mask on the picture.
• Quantity of mask detected
• threshold configurated to evaluate that picture.

You can configure the threshold value to detect with more or less precision.

🎯 Purpose

This project was done for educational purposes, using Custom Vision to train the model with 629 facemask examples and node-red programming tool. You can test it using:

• Basic flow using an input file.

• Advanced usage with sp32-cam and RPI how Gateway (IoT Implementation).