anpr_scanner_flutter 0.0.1

A Flutter package for Automatic Number Plate Recognition (ANPR) using TFLite models.

0.0.1

• Initial release of the anpr_scanner_flutter package.

Core Features

• Two-Stage ML Pipeline: Integrated TFLite support for sequential license plate detection and optical character recognition (OCR).
• Drop-in Scanner Widget: Added AnprScannerWidget, a fully self-contained UI that manages camera permissions, model loading, real-time scanning, and result presentation.
• Low-Level API: Added LicensePlateDetector to allow headless, direct processing of img.Image objects from any source (camera, gallery, or file system).
• Smart OCR Parsing: Built-in post-processing that sorts characters left-to-right and intelligently splits plate codes and numbers using regex and largest-gap heuristics.

Performance & Optimization

• Hybrid Hardware Acceleration: Implemented an automated delegate strategy that routes the heavier detection model to the GPU while keeping OCR on the CPU to prevent native driver crashes.
• Optimized Image Conversion: Added CameraImageConverterOptimized utilizing pre-computed YUV lookup tables and isolate offloading for fast YUV-to-RGB conversion.
• Async Model Caching: Introduced ModelServiceManagerOptimized as a singleton to handle asynchronous model loading and memory management.

Real-time Scanning & Stability Tracking

• Auto-Capture Gates: Introduced CaptureQualityConfig to ensure high-quality crops by enforcing configurable stability thresholds (minimum confidence, box drift, area change, and consecutive stable frames).
• Frame Pacing: Added RealtimeConfig to control global tuning constants like frame downsample factors and minimum interval processing times.
• Frame Processing Pipeline: Added DetectionOnlyProcessor to manage the continuous stream of camera frames efficiently.

UI & UX Components

• Interactive Camera View: Integrated pinch-to-zoom, tap-to-focus, and torch toggle functionalities directly into the scanner widget.
• Visual Feedback: Added DetectionOverlay for smooth, animated bounding boxes during the scanning phase.
• State Overlays: Included ModelLoadingOverlay for initialization states and ScanResultOverlay (with flash animation) for the final capture transition.
• Permission Handling: Automated camera permission requests and state management within the high-level widget.