flutter_capture_vision

A cross-platform Flutter plugin for barcode reading, MRZ (passport/ID) recognition, and document boundary detection, powered by Dynamsoft Capture Vision. Supports Android, iOS, Web, Windows, Linux, and macOS.

The plugin wraps the Dynamsoft Capture Vision SDK of each platform behind one stable Dart API. It accepts image files or raw pixel buffers, so it composes with any camera plugin — for example flutter_lite_camera for live preview and frame capture.

Getting Started

1. Install the Package

dependencies:
  flutter_capture_vision: ^0.1.0

The Android, iOS, Web, Windows, macOS, and Linux implementations are endorsed federated packages and are selected automatically — depend only on flutter_capture_vision.

2. Obtain a License Key

A valid license is required to activate barcode, MRZ, and document detection.

3. Initialize the SDK

import 'package:flutter_capture_vision/flutter_capture_vision.dart';

final vision = FlutterCaptureVision();
await vision.initialize(licenseKey: 'YOUR-LICENSE-KEY');

Usage

Capture from an Image File

final result = await vision.captureFile(
  '/path/to/image.jpg',
  const CaptureVisionRequest.forTasks({VisionTask.barcode}),
);

for (final barcode in result.barcodes) {
  print('${barcode.format}: ${barcode.text}');
}

Capture from a Camera Buffer

final result = await vision.captureBuffer(
  VisionImageBuffer(
    bytes: rgbBytes,       // Uint8List
    width: width,
    height: height,
    stride: width * 3,
    pixelFormat: VisionPixelFormat.rgb888,
    rotation: VisionRotation.degrees0,
  ),
  const CaptureVisionRequest.forTasks({
    VisionTask.barcode,
    VisionTask.mrz,
    VisionTask.documentDetection,
  }),
);

CaptureVisionResult always returns non-null lists; "nothing found" is an empty list. Failures (license, invalid input, native SDK errors) throw CaptureVisionException(code, message) with a stable machine-readable code.

Vision Tasks

Task	Built-in template	Results
VisionTask.barcode	ReadBarcodes_Default	result.barcodes
VisionTask.mrz	ReadPassportAndId	result.mrzResults
VisionTask.documentDetection	DetectDocumentBoundaries_Default	result.documentDetections

MRZ results carry the composed rawText, the parsed fields (passport number, date of birth, …), and the documentType (MRTD_TD3_PASSPORT, MRTD_TD1_ID, …). Passports and ID cards need reasonably sharp, high-resolution input — request at least 1080p from the camera for MRZ scanning.

Custom Settings

Advanced users can replace the built-in templates with a Dynamsoft Capture Vision template JSON:

final templates = await vision.initSettings(settingsJson);   // returns parsed template names
final result = await vision.captureFile(
  path,
  CaptureVisionRequest.namedTemplate(templates.first.name),
);
await vision.resetSettings();                                // restore factory defaults

Platform Notes

• Android: minSdkVersion 21.
• iOS: requires iOS 13+; add NSCameraUsageDescription when using a camera plugin.
• Web: serve over HTTPS or localhost. The JS/WASM/models bundle is self-hosted as package assets — no CDN <script> tag is needed.
• Windows/Linux/macOS: the released app bundle carries the DCV libraries and resources; see each platform package README for details.

Examples

Each package ships a runnable example app:

cd example
flutter run                 # Android/iOS desktop or device
flutter run -d chrome       # Web
flutter run -d windows      # Windows/Linux/macOS

The facade example is the full six-platform demo with live camera capture, annotation overlay, and document export.

License

See LICENSE.