Facial Expression Recognition for your Flutter application in a few lines of code.
Mojo Facial Expression Recognition provides :
❤️ 3 emotions :
- amusement
- confusion
- surprise
🎉 3 social cues :
- attention
- engagement
- interaction status
💡 Open Source API ⚡️ Real-time 🔐 private-by-design
No images are never sent to the the cloud.
Handle your user's emotional reactions to build on top of your amazing app
Get your free trial API key.
Add this line like this to your project's pubspec.yaml and run flutter pub get:
dependencies:
mojo_perception: ^2.0.0
Add row to the ios/Runner/Info.plist with the key Privacy - Camera Usage Description and a usage description :
<key>NSCameraUsageDescription</key>
<string>Can I use the camera please?</string>
In the pubspec.yaml add the following line:
assets:
- packages/mojo_perception/assets/face_landmark_with_attention.tflite
- packages/mojo_perception/assets/face_detection_short_range.tflite
Run
flutter pub get
-
Download TensorFlowLiteC.framework.zip (TensorFlowLiteC framework with Mediapipe special ops).
-
Unzip
TensorFlowLiteC.frameworkin the pub-cache folder of tflite_flutter package: Pub-Cache folder location:
- ~/.pub-cache/hosted/pub.dartlang.org/tflite_flutter-0.9.0/ios/ (Linux/ Mac)
- %LOCALAPPDATA%\Pub\Cache\hosted\pub.dartlang.org\tflite_flutter-0.9.0\ios\ (Windows)
-
Download jniLibs.zip.
-
Extract in your project path :
android > app > src > main
In the pubspec.yaml add the following line:
assets:
- packages/mojo_perception/assets/face_landmark_with_attention.tflite
- packages/mojo_perception/assets/face_detection_short_range.tflite
Run
flutter pub get
If you encounter an error concerning minSdkVersion:
-
Go to
android > src > app > build.gradle -
Find line
minSdkVersion flutter.minSdkVersion -
Change
flutter.minSdkVersionby 21
Import it:
import 'package:mojo_perception/mojo_perception.dart';
Create an object MojoPerceptionAPI:
MojoPerceptionAPI mojoPerceptionApi = MojoPerceptionAPI(
'<auth_token>',
'<host>',
'<port>',
'<user_namespace>');
Please note that the user token 'auth_token' is different from the API Key.
🙏 We take a particular care with publishable user token for applications. See below for more info and best practices.
To make things safe and easy, we have a REST API ready to create user tokens on-the-fly.
When using the API for a web application, you should implement a backend function that will get a token for each user. We recommend to use tokens with expiration date.
To do so, you can use the REST API of Mojo Perception API.
Below, an example using curl :
In your terminal, replace <YOUR_API_KEY_HERE> by one of your API Keys and run :
curl -X PUT -i -H 'Authorization: <YOUR_API_KEY_HERE>' -d '{"datetime": "2022-01-01T00:00:00.000000","expiration":360}' -H 'Content-Type: application/json' https://api.mojo.ai/mojo_perception_api/user
This will give you something like this, and note that we have set expiration: 360, for a 60 x 60 seconds duration period :
HTTP/2 200
server: nginx
date: Tue, 18 Jan 2022 16:53:54 GMT
content-type: application/json
content-length: 350
{"user_namespace": "a5fa97ded6584cb4a7ff3933aa66025c", "auth_token": "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpYXQiOjE2NDI1MjQ4MzQsInN1YiI6ImE1ZmE5N2RlLWQ2NTgtNGNiNC1hN2ZmLTM5MzliYTY2MDI1YyIsImV4cCI6MTY0MjUyNDg0NH0.7FuLJ6Hmozi2DbX9zooVxYvnp_f91H4vzodstDZbLzI", "host_name": "socket.mojo.ai", "port": "443"}
You can use the auth_token and given host_name,port and user_namespace to configure your API endpoint in the API, with the setOptions() method.
💡 Have a look to the tutorials section
If you face a "JsonWebTokenError", maybe that's because of the expiration. You can try to increase the user token duration to match your need. Default value of 360 seconds might be too short.
Special thanks to JaeHeee for this project that inspired us implementing mediapipe part.