vision_ai_models #

Bundled hand-gesture and face-emotion ML models for the vision_ai plugin.

As of vision_ai 0.3.0 the core plugin bundles no models — it loads each model from a file path supplied at runtime. This keeps an app's footprint to only the models it actually uses. vision_ai_models carries the hand/face models, copies them into app storage on first use, and hands back the resolved paths as a ModelPaths.

Add it whenever you use HandConfig or FaceConfig(detectEmotion: true). If you only do animal detection, you don't need this package — use vision_ai_animals instead.

Installation #

dependencies:
  vision_ai: ^0.4.0
  vision_ai_models: ^0.1.0

Usage #

import 'package:vision_ai/vision_ai.dart';
import 'package:vision_ai_models/vision_ai_models.dart';

// Copy bundled models to app storage once and get their paths.
final models = await VisionAiModels.ensureLoaded();

final vision = VisionAi(
  hand: const HandConfig(),
  face: const FaceConfig(detectEmotion: true),
  models: models,
);

final textureId = await vision.start();

ensureLoaded() is safe to call on every launch — it skips the copy when a file of the matching size already exists.

What it bundles #

Both are MediaPipe/TFLite models loaded by the core engine. The copy targets getApplicationSupportDirectory()/vision_ai_models/ and works on Android + iOS.

For animal models, see vision_ai_animals. To supply your own model paths without this package, build a ModelPaths directly.

License #

Apache License 2.0. The bundled models are distributed by Google under the Apache 2.0 license.