https://github.com/jsflo/facedetect

https://github.com/jsflo/facedetect

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/jsflo/facedetect
• Owner: JsFlo
• Created: 2018-05-22T22:44:55.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-09-07T02:33:31.000Z (over 6 years ago)
• Last Synced: 2024-04-20T17:21:01.198Z (8 months ago)
• Language: Kotlin
• Size: 5.65 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# FaceDetect

Android app that uses Firebase's ML Kit to detect faces and then can also provide

emotion predictions by using a custom trained model.

More information about the ML Kit's implementation can be found in [this article](https://medium.com/rocket-fuel/quick-look-face-detection-on-android-using-ml-kit-7041de41df4c) I wrote going over ML Kit's face detection.

### Note when running the for the first time

* ML Kit downloads the model for local inference the first time you use it

  * After running for the first time (after enabling permissions) you may have to wait some time

* Handling permissions have not been handled yet

  * Current Workaround: enable permissions for the app in the Android OS settings

* Emotion Recognition is pointing to a local server using ngrok

  * You would need to set your own instance and change url

## ML Kit's Face Detector

Wraps the `FiresbaseVisionFaceDetector` from the new `ml.vision` package in a `Detector` object from the

 old `gms.vision` package so we can use the `CameraSource` and capture a controlled stream of frames and predictions.



 ## Custom Emotion Detection

 

 You can setup an instance of a server like I have [here](https://github.com/JsFlo/EmotionRecServer) and update the static variable in the `RetrofitNetwork.kt` file.

 ```java

 val BASE_URL = "https://xxxxxx.ngrok.io"

 ```

*I use [ngrok](https://ngrok.com/) to expose the endpoints running from a local server*