EyaHaddad/tsl-mobile-app

Monolithic Flutter mobile application for offline Tunisian Sign Language recognition. # tsl_mobile_app Flutter mobile app for Tunisian sign language recognition with on-device inference, local history persistence, and text to speech playback ## Current Stack - camera - tflite_flutter - tflite_flutter_helper - flutter_tts - permission_handler - shared_preferences - isar - isar_flutter_libs - freezed and json_serializable - flutter_svg - intl ## Main Features - Camera capture flow for recognition - On-device TensorFlow Lite inference - Local history saved in Isar - Favorites support in history - Auto-delete for expired non-favorite records based on settings - On-demand text to speech from result and history detail screens - Persistent storage settings with SharedPreferences ## Project Structure ```text lib/ ├── main.dart ├── core/ │ ├── constants/ │ ├── database/ │ │ └── isar_service.dart │ ├── routes/ │ ├── services/ │ │ ├── history_retention_service.dart │ │ └── text_to_speech_service.dart │ ├── theme/ │ └── utils/ ├── features/ │ ├── camera/ │ ├── history/ │ │ ├── models/ │ │ │ └── history_record.dart │ │ ├── screens/ │ │ │ ├── history_screen.dart │ │ │ └── item_history_screen.dart │ │ └── services/ │ │ └── history_storage.dart │ ├── home/ │ ├── recognition/ │ │ ├── managers/ │ │ ├── models/ │ │ ├── screens/ │ │ │ └── result_screen.dart │ │ └── services/ │ └── settings/ │ ├── models/ │ │ └── settings_model.dart │ ├── screens/ │ │ └── settings_screen.dart │ └── services/ │ └── settings_service.dart └── shared/ ├── providers/ └── widgets/ ``` ## Recognition Pipeline Camera -> MediaPipe landmarks -> sequence buffer -> LSTM inference -> predicted text ## Audio Behavior - Audio is practical and on-demand - User taps play or convert to voice to run TTS - Language, rate, and pitch are loaded from saved settings ## History Retention Rules - Favorites are persistent - Non-favorite records can be removed automa …