This is an implementation of the facerecognition.js module on NodeJS.
Pre-requisite
Download the face-recognition npm module and follow the build instructions according to your system
Usage
Next, git clone the project, and do npm install. If you have built performed the previous build step correctly, this step should go off without a hitch.
Then, npm start gets the server started on http://localhost:3000
Image Storage
Some sample faces are provided in https://github.com/hhk998402/FaceRecogition-NodeJS/blob/master/routes/faces All newly added faces are stored here. Store the images as PNG Files only. Also, name the files in the following way: -
firstName-middleName-lastName_anyString.png
Make sure the fileName doesn't consist of any spaces or any other special character apart from underscore. Use hyphens in place of spaces for longer names.
Routes and their Usage
-
http://localhost:3000/capture (GET) - This is the main page, wherein you can test and see the predictions as well as train new faces. To perform facial recognition against the images that have already been trained, go to
http://localhost:3000/capture/and click on the camera icon in the capture window. Wait for about three seconds and the prediction will be shown above with the closeness to the answer. Remember, here the lower the distance, the better the prediction.To take another image, click the trash icon to re-engage the video feed, and take another snap.
Here is a simple demonstration:
Now, to train new faces, first type your name in the input box below the capture window. Don't use underscores or spaces. Use hyphens in place of spaces. After entering a valid name, click the "Click here to train" button alongside. Make sure you are prominently visible in the capture box. Wait for the training completed message to come before adding a new image.
-
http://localhost:3000/predict (POST) - It is accessed through an AJAX POST request after pressing the camera icon in the capture box. It is used to return the predicted face in the image that has been sent to it. The image is passed to the route in base64 format.
-
http://localhost:3000/train (POST) - It is accessed through an AJAX POST request after pressing the Train Button. It is used to add faces to both existing and new users..