A Flutter plugin providing fast, flexible TensorFlow Lite inference with multi-platform delegate support.
TensorFlow Lite Flutter plugin is a Flutter package that enables developers to run TensorFlow Lite machine learning models directly within Flutter applications. It provides a flexible and fast interface for model inference, supporting hardware acceleration across Android, iOS, and desktop platforms. The plugin binds directly to the TensorFlow Lite C API, ensuring low-latency performance similar to native implementations.
Flutter developers who need to integrate on-device machine learning models into their mobile or desktop applications, particularly those requiring cross-platform support and hardware-accelerated inference.
Developers choose this plugin for its near-native inference speeds, multi-platform delegate support (NNAPI, GPU, Metal, CoreML), and API consistency with TensorFlow Lite's official Java and Swift SDKs, making it easier to port existing ML workflows to Flutter.
TensorFlow Lite Flutter Plugin
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Supports delegates like NNAPI, GPU, Metal, and CoreML across Android, iOS, and desktop, enabling optimized inference speeds as highlighted in the key features.
Direct binding to TensorFlow Lite C API ensures low-latency inference, with speeds close to native apps, per the README's efficiency claims.
Runs inference in separate isolates to prevent jank, a key feature for maintaining smooth Flutter app performance during ML operations.
Mirrors TensorFlow Lite's Java and Swift APIs, making it easier for developers familiar with those SDKs to transition and use the plugin.
Requires manual steps like downloading binaries, running platform-specific scripts, and handling framework placements, which can be error-prone and time-consuming.
The project is under active migration to a new repository with admitted broken elements, as noted in the announcement, making it risky for stable deployments.
Only compatible with TensorFlow Lite models, excluding other ML frameworks like PyTorch Mobile or ONNX Runtime that might be needed for specific use cases.