This is an example of how to use the NudeNet model in a flutter app.
The pre-trained model can be found here https://github.com/notAI-tech/NudeNet
The demo can be found here : https://drive.google.com/file/d/101SZligJpFSl72KTc0fBPJfL1ardb_jZ/view?usp=sharing
Steps
-
Install flutter ( flutter channel stable): https://flutter.dev/docs/get-started/install?gclid=CjwKCAiAtej9BRAvEiwA0UAWXsbdQkuR73-UtVp58l50mlCbb38_dR7qlTUazhnaSgmNuL82lsF2xBoCRcoQAvD_BwE&gclsrc=aw.ds
-
Create a new flutter project: https://flutter.dev/docs/get-started/codelab#step-1-create-the-starter-flutter-app
-
Download the classifier.tflite file from https://github.com/notAI-tech/NudeNet/releases
-
Create a folder "assets" in the root of your project directory
-
Create labels.txt file under assets/ and add these lines :
0 safe1 unsafe -
Copy classifier.tflite into the assets folder
-
Go to pubspec.yaml file and add these dependencies :
tflite: any
image_picker: ^0.6.2+3
-
Run:
flutter pub get -
Go to android\app\build.gradle and add theses lines below
release {}:aaptOptions{noCompress 'tflite'noCompress 'lite'
}
-
In the same file (android\app\build.gradle) check if the minSdkVersion in
defaultConfig{}is 19, if not set it to 19 (minSdkVersion 19) -
Restart your IDE (vscode, android studio, etc)
-
Delete all the code in lib/main.dart and paste my main.dart code into your main.dart
-
Save and launch the app:
flutter run
Test it using nude photos and decent photos on your android smartphone or in the emulator.
Note: - The model is not 100% accurate, sometimes it detects a banana as an unsafe picture. I tested it on android device only
Good luck!!!
The demo can be found here: https://drive.google.com/file/d/101SZligJpFSl72KTc0fBPJfL1ardb_jZ/view?usp=sharing
Credits : https://github.com/notAI-tech/NudeNet