How ananyajackson11/clearsteer works
Project type and purpose
• ClearSteer is a cross-platform (mobile & web) Flutter application.
• Its core purpose is to capture camera input, perform on-device ML inference, provide text-to-speech feedback, and persist user data via Firebase.
File structure
• Top-level “android”, “ios”, “linux”, “macos”, “windows”, “web” folders: each hosts platform-specific Flutter embedding code.
• “lib/” contains Dart source files: UI pages (login, signup, drive, emergency, statistics, profile), plus detect.dart (ML wrapper).
• “assets/50epoch_float32.tflite” – the TFLite model.
• “pubspec.yaml” – Flutter dependencies (Firebase, tflite_flutter, flutter_tts, camera, image_picker).
• “test/” – widget tests.
• Platform folders include plugin registrants and runner code for desktop/embedded targets.
README
• Tech stack: Flutter UI, Firebase Auth & Firestore, TFLite, camera/TTS utilities, HTTP for external calls.
• Features: camera integration, real-time TTS, on-device ML, Firestore persistence, cross-platform.
Main components and interactions
a. Components
1. Flutter UI Layer (lib/*.dart pages)
2. Device Services
– Camera plugin
– Image Picker plugin
– flutter_tts plugin
3. ML Inference Engine
– TFLite runtime (tflite_flutter)
– Local model file (assets/50epoch_float32.tflite)
4. Backend Services (Firebase)
– Authentication (Firebase Auth)
– Data Storage (Cloud Firestore, optionally Storage)
5. Network Layer
– HTTP client (for optional REST calls)
b. Relationships & Data Flow
– UI captures camera frames → passes to ML Inference → outputs detection results.
– Detection results → fed to flutter_tts → audio feedback to user.
– User actions (login/signup) → Firebase Auth → returns user token.
– Authenticated UI → reads/writes user data to Firestore.
– Image uploads/read from Firestore Storage (if used).
– HTTP requests (if any) go out via Network Layer.
c. Architectural patterns & principles
– Layered architecture: Presentation (UI) → Services (Camera, TTS, HTTP, ML) → Backend (Firebase).
– Plugin-based modularity: each platform service is a Flutter plugin.
– Cross-platform runner host code isolates embedder details (Android/iOS/Linux/macOS/Windows).
d. Key technologies
– Flutter framework (Dart)
– TFLite Flutter plugin
– flutter_tts
– firebase_auth, cloud_firestore
Detailed capture of essential elements
• Platform boundaries: a container for Flutter host on Android/iOS/Desktop/Web.
• Model artifact as a runtime dependency.
• Real-time loop: camera frame → ML → TTS.