advanced_image_processing_toolkit 0.2.0

Flutter image processing toolkit with filters, geometric transforms, watermarking, ML-powered detection (objects, faces, text, pose), and AR hooks.

Advanced Image Processing Toolkit

Filters, geometric transforms, watermarking, ML-powered detection, and AR hooks for Flutter.

A Flutter package that combines:

• Filters — grayscale, blur, brightness, sepia, invert, vignette, watercolor, oil-painting, contrast, saturation
• Geometric transforms — resize, rotate, crop, flip
• Watermarking — image overlay compositing
• ML detection — objects, faces, text (OCR), human pose via Google ML Kit
• Augmented reality — method-channel bindings (BYO ARKit/ARCore plugins)

Installation

dependencies:
  advanced_image_processing_toolkit: ^0.2.0

Filters and transforms run on Android, iOS, and Web. ML Kit detection requires Android or iOS (Google ML Kit is mobile-only).

Optional: AR support

AR is not bundled — it would balloon the install for every consumer. To use AugmentedReality.*, add the platform AR plugins yourself:

dependencies:
  arkit_plugin: ^1.1.0          # iOS
  arcore_flutter_plugin: ^0.1.0 # Android (community-maintained)

Quick start

import 'package:advanced_image_processing_toolkit/advanced_image_processing_toolkit.dart';

await AdvancedImageProcessingToolkit.initialize(
  enableObjectDetection: true,
  enableAR: false,
);

Filters

All filter methods are static on ImageFilters, take a Uint8List, and return a Future<Uint8List>.

// Basic
final gray   = await ImageFilters.applyGrayscale(bytes);
final blur   = await ImageFilters.applyBlur(bytes, 5.0);        // sigma
final sepia  = await ImageFilters.applySepia(bytes);
final invert = await ImageFilters.applyInvert(bytes);

// Brightness (delta, -1.0 .. 1.0)
final brighter = await ImageFilters.adjustBrightness(bytes, 0.5);
final darker   = await ImageFilters.adjustBrightness(bytes, -0.5);

// Contrast / saturation (multiplier, 1.0 is neutral)
final contrasty = await ImageFilters.adjustContrast(bytes, 1.5);
final vivid     = await ImageFilters.adjustSaturation(bytes, 1.2);

Artistic effects

final vignette = await ImageFilters.applyVignette(
  bytes,
  intensity: 0.5,  // 0..1
  radius: 0.5,     // 0..1
);

final watercolor = await ImageFilters.applyWatercolor(
  bytes,
  radius: 5,       // blur radius
  intensity: 0.6,
);

final oil = await ImageFilters.applyOilPainting(
  bytes,
  radius: 4,       // brush size
  levels: 20,      // colour quantisation
);

Geometric transforms

final resized = await ImageFilters.applyResize(bytes, width: 800); // height auto
final rotated = await ImageFilters.applyRotate(bytes, 90.0);
final cropped = await ImageFilters.applyCrop(bytes, 100, 100, 400, 400);
final flipped = await ImageFilters.applyFlip(bytes, horizontal: true);

Watermarking

final watermarked = await ImageFilters.applyWatermark(
  baseBytes,
  logoBytes,
  x: 50,
  y: 50,
  opacity: 0.5,
);

Filter chains

Filters compose by simply passing the output of one into the next:

var out = await ImageFilters.applyGrayscale(bytes);
out     = await ImageFilters.adjustContrast(out, 1.2);
out     = await ImageFilters.applyVignette(out, intensity: 0.3);

ML Kit detection

ObjectRecognition.detectObjectsFromPath runs object detection, face detection, pose estimation, and text recognition in parallel and returns a unified list of DetectedObject. Filter by label to slice the kind of detection you want.

final detections =
    await ObjectRecognition.detectObjectsFromPath('/path/to/photo.jpg');

for (final d in detections) {
  switch (d.label) {
    case 'Face':
      final smiling = d.additionalData?['smilingProbability'] as double?;
      print('Face (smile=$smiling) at ${d.boundingBox}');
    case 'Text':
      print('OCR: ${d.additionalData?['text']}');
    case 'Person':
      print('Pose with ${(d.additionalData?['landmarks'] as List).length} landmarks');
    default:
      print('${d.label} (${(d.confidence * 100).toStringAsFixed(0)}%) at ${d.boundingBox}');
  }
}

ML Kit requires a file path or accurate InputImageMetadata. If you have raw bytes and you know the image's width/height/stride/format:

final detections = await ObjectRecognition.detectObjectsFromBytes(
  bytes,
  InputImageMetadata(
    size: Size(width.toDouble(), height.toDouble()),
    rotation: InputImageRotation.rotation0deg,
    format: InputImageFormat.bgra8888,
    bytesPerRow: width * 4,
  ),
);

Call ObjectRecognition.dispose() when your app is shutting down to release the underlying ML Kit detectors.

Augmented reality

if (AugmentedReality.isARSupported()) {
  await AugmentedReality.startARSession();
  await AugmentedReality.placeModel(
    modelPath: 'assets/models/chair.glb',
    position: [0, 0, -1],   // x, y, z
    scale: 1.0,
    rotation: [0, 0, 0],
  );
  // ...
  await AugmentedReality.stopARSession();
}

isARSupported() only checks the OS — it does not verify the consumer app has actually added arkit_plugin / arcore_flutter_plugin. AR calls return false and log a warning if the channel isn't wired.

Platform support

Platform	Filters & transforms	ML detection	AR
Android	Yes	Yes	Yes (with ARCore plugin)
iOS	Yes	Yes	Yes (with ARKit plugin)
Web	Yes	No (throws UnsupportedError)	No
Desktop	Yes	No	No

Performance tips

Run heavy filters off the UI thread with compute:

final out = await compute(_grayscale, bytes);
Uint8List _grayscale(Uint8List b) =>
    /* call ImageFilters.applyGrayscale here */;

Resize first, then process — Gaussian blur and oil-painting are quadratic in pixel count:

final small = await ImageFilters.applyResize(huge, width: 1920);
final out   = await ImageFilters.applyBlur(small, 5);

API reference

See API_REFERENCE.md.

Author

Godfrey Lebo — Fullstack Developer & Technical PM

• Email: [e-mail]
• LinkedIn: godfreylebo
• Portfolio: godfreylebo.dev
• GitHub: @emorilebo

License

MIT — see LICENSE.