hand_detection
4.3.2, released 7 days ago by hugo.ml
On‑device hand detection, landmarks and gesture recognition using LiteRT models.
What it does
The package runs MediaPipe hand detection locally, returning 21‑point 3D landmarks, handedness and a set of predefined gestures. It works on mobile, desktop and web without any remote API.
AI-written summary of version 4.3.1, based on the package's README and metadata. Report an inaccuracy
Use it when
- You need offline hand tracking on multiple platforms
- You require 3D hand landmarks and built‑in gesture classification
- You want GPU/NPU acceleration where available
Skip it when
- You prefer a pure Dart solution without native OpenCV dependencies
- You target very old devices or browsers without WebGPU support
Strengths
- Fully offline, no network calls
- Cross‑platform: Android, iOS, macOS, Windows, Linux, Web
- Supports GPU/NPU via LiteRT for better performance
- Provides ready‑made gesture labels
Weaknesses
- Requires native libraries (opencv_dart, dartcv4) and specific Flutter/Dart versions
- Setup can be complex (hooks, iOS deployment target, native linking)
- Web support limited to WebGPU
Setup
- Kind
- Plugin (native code)
- Effort
- High
- Code generation
- No
- Direct dependencies
- 6
Requires Dart 3.10+ and Flutter 3.47.5+. Add hooks user_defines for iOS deployment target (>=15.0). Ensure native dependencies (opencv_dart, dartcv4) are compiled for each platform. Include camera_desktop for desktop camera use. May need to run flutter clean after changes.
import 'dart:io';
import 'package:hand_detection/hand_detection.dart';
Future<void> main() async {
final detector = await HandDetector.create();
final bytes = await File('path/to/image.jpg').readAsBytes();
final hands = await detector.detect(bytes);
for (var hand in hands) {
print('Handedness: ${hand.handedness}');
if (hand.hasLandmarks) {
final wrist = hand.getLandmark(HandLandmarkType.wrist);
print('Wrist: (${wrist?.x}, ${wrist?.y})');
}
}
await detector.dispose();
}
Alternatives
Packages that do the same job and could replace hand_detection.
- tflite_flutter
Provides a lower‑level TFLite API without the heavy OpenCV and LiteRT stack, suitable for simpler models or when you want full control over inference.
68Compare - google_mlkit_commons
Offers Google‑backed ML Kit APIs with easier setup on mobile, though it relies on Google services and may not support desktop/web.
89Compare
Works with
Packages commonly used together with hand_detection: ones it builds on, ones built on it, and its own family.
- webhand_detection depends on itLightweight bindings for browser APIs via JS interop
- hookshand_detection depends on itDart API for the JSON-based protocol for package build and link hooks.
- dartcv4hand_detection depends on itOpenCV4 bindings for Dart/Flutter using ffi for native computer‑vision functions.
- flutter_literthand_detection depends on itLiteRT (formerly TensorFlow Lite) Flutter plugin. Drop-in on-device ML inference with bundled native libraries for supported native platforms and web runtimes.
- opencv_darthand_detection depends on itFlutter bindings for OpenCV4 using dart:ffi.
- face_detection_tfliteSame familyAdvanced face & landmark detection, embedding and segmentation using on-device LiteRT (formerly TensorFlow Lite) models.
- pose_detectionSame familyPose, person and landmark detection using on-device LiteRT (formerly TensorFlow Lite) models.