Only run face recognition on trigger from external notification. (#156)

CHANGELOG.MD

@@ -1,3 +1,10 @@
+# 2.0.2 (2024-11-12)
+- Added feature to conditionally trigger face recognition to start and stop via a Magic Mirror notification. Use of this feature is optional. Use for example with MMM-Pir to turn off face recognition when monitor is off, and back on when monitor is on.
+  - node_helper sends commands to the python child process via its std input upon receiving an external notfication. 
+  - CPU utilization of python process is substatially reduced while face recognition is paused. RPi power consumption also reduced
+- Added SIGTERM signal handling to recognition.py so python process reliably stops when Magic Mirror application is terminated
+- Improved recognition.py loop timing so the target interval specified by `checkInterval` is better enforced by the main loop
+
 # 2.0.1 (2023-11-01)
 
 - After a lot of install issues. We removed `pipenv` again and changed the install-instructions

MMM-Face-Reco-DNN.js

@@ -71,6 +71,12 @@ Module.register('MMM-Face-Reco-DNN', {
     outputmm: 0,
     // turn on extra debugging 0=no, 1=yes
     debug: 0,
+    // If specified, conditionally run face recognition when a notification with a name as specified by this 
+    // option is received with a boolean true/false payload. A True payload activates face recognition, False 
+    // deactivates it. An empty string here disables this functionality so face recognition always runs. Use for 
+    // exmaple with with MMM-Pir with notification name 'MMM_PIR-SCREEN_POWERSTATUS' to only run face 
+    // recognition when screen is on.
+    external_trigger_notification: '',
   },
 
   timouts: {},
@@ -466,6 +472,13 @@ Module.register('MMM-Face-Reco-DNN', {
       Log.log(this.name + ' get logged in users ' + this.users);
       this.sendNotification('LOGGED_IN_USERS', this.users);
     }
+
+    // If an external trigger notification name is specified, forward
+    // notification payload received on that notification name.
+    if (this.config.external_trigger_notification !== "" &&
+        notification === this.config.external_trigger_notification) {
+      this.sendSocketNotification(notification, payload);
+    }
   },
 
   getDom: function () {

README.md

@@ -21,9 +21,12 @@ This module works in the background, and so no screenshots are available.
 
 The installation is much more simplified now. You can run `npm ci` to install all the node packages and after that it will install all packages with pip. This can take a while to compile the whole opencv stuff.
 
-Be sure that your raspberry pi has enough cooling to do this job, it will be heavy used.
+Be sure that your raspberry pi has enough cooling to do this job, it will be heavily used. It is also 
+likely necessary to increase the available SWAP space for opencv and dlib to succesfully compile. SWAP 
+can be increased by editing the file: `sudo nano /etc/dphys-swapfile` and changing
+the value of `CONF_SWAPSIZE`. 1024 MB was enough to work on a RPi 4 (i.e. `CONF_SWAPSIZE=1024`). 
 
-But that this is working you need installed system wide tools before:
+For the installation procedure to work, you first need the following installed beforehand system wide:
 
 - node with nvm
   - [install instructions](https://github.com/nvm-sh/nvm#installing-and-updating)
@@ -35,15 +38,16 @@ But that this is working you need installed system wide tools before:
   - `sudo apt install libcap-dev`
 - install python dependencies
   - If you working with Bookworm you need to create first an virtual environment, please have a look in the next chapter
-  - `pip install face-recognition numpy dlib picamera2 opencv-python`
+  - Numpy versions 2.0+ do not seem to work with this module
+  - `pip install face-recognition numpy==1.26.4 dlib picamera2 opencv-python`
 
 ### Some additional steps for Bookworm and above to run it with an virtual environment
 
 If you want/need to install it with an virtual environment, you need to do following steps before you install the packages with pip
 
 - create environment with `python3 -m venv ~/python-facereco`
 - activate environment with `source ~/python-facereco/bin/activate`
-- install pip packages with `pip install face-recognition numpy dlib picamera2 opencv-python`
+- install pip packages with `pip install face-recognition numpy==1.26.4 dlib picamera2 opencv-python`
 - Because some libraries uses global installed libs which are not available with pip, you need to change the config of your virtual environment
   - `nano ~/python-facereco/pyvenv.cfg`
   - Change line `include-system-site-packages = false` to `include-system-site-packages = true`
@@ -166,6 +170,12 @@ To setup the module in MagicMirror², add the following section to the `config.j
       outputmm: 0,
       // turn on extra debugging 0=no, 1=yes
       debug: 0,
+      // If specified, conditionally run face recognition when a notification with a name as specified by this 
+      // option is received with a boolean true/false payload. A True payload activates face recognition, False 
+      // deactivates it. An empty string here disables this functionality so face recognition always runs. Use for 
+      // exmaple with with MMM-Pir with notification name 'MMM_PIR-SCREEN_POWERSTATUS' to only run face 
+      // recognition when screen is on.
+      external_trigger_notification: '',
     }
 }
 ```

node_helper.js

@@ -21,6 +21,7 @@ module.exports = NodeHelper.create({
     const extendedDataset = this.config.extendDataset ? 'True' : 'False';
     const options = {
       mode: 'json',
+      pythonOptions: ['-u'], // Immediately flush buffer for std out/in monitoring/writing to work
       stderrParser: line => JSON.stringify(line),
       args: [
         '--cascade=' + this.config.cascade,
@@ -38,6 +39,7 @@ module.exports = NodeHelper.create({
         '--contrast=' + this.config.contrast,
         '--resolution=' + this.config.resolution,
         '--processWidth=' + this.config.processWidth,
+        '--run-only-on-notification=' + (this.config.external_trigger_notification !== '' ? '1' : '0'),
       ],
     };
 
@@ -81,22 +83,26 @@ module.exports = NodeHelper.create({
       }
     });
 
-    // Shutdown node helper
-    self.pyshell.end(function (err) {
-      if (err) throw err;
-      console.log('[' + self.name + '] ' + 'finished running...');
-    });
-
     onExit(function (_code, _signal) {
       self.destroy();
     });
   },
 
+  send_python_cmd: function (cmd) {
+    this.pyshell.send(cmd);
+  },
+
   python_stop: function () {
     this.destroy();
   },
 
   destroy: function () {
+    const self = this
+    this.pyshell.end(function (err) {
+      if (err) throw err;
+      console.log('[' + self.name + '] ' + 'finished running...');
+    });
+
     console.log('[' + this.name + '] ' + 'Terminate python');
     this.pyshell.childProcess.kill();
   },
@@ -112,6 +118,16 @@ module.exports = NodeHelper.create({
         this.python_start();
       }
     }
+
+    // Notification for triggering face recognition received. Only send to python subprocess
+    // if it has been started
+    if (notification === this.config.external_trigger_notification && pythonStarted) {
+      if (payload === true) {
+        this.send_python_cmd('start');
+      } else {
+        this.send_python_cmd('stop');
+      }
+    }
   },
 
   stop: function () {