android-on-device-ml

What problem does it solve?

It helps you add on-device machine learning to Android apps so you can recognize text, detect objects, estimate poses, and run custom TensorFlow Lite inference without needing a server.

Core Features & Use Cases

• ML Kit integration for common vision tasks like OCR, face detection, and object detection with minimal model management.
• TensorFlow Lite custom model inference with control over tensor I/O, interpreter lifecycle, and performance delegates like GPU/NNAPI.
• MediaPipe Tasks API for higher-level vision/pose/gesture workflows with GPU acceleration and task bundles packaged in app assets.
• Use case example: In a CameraX image analysis pipeline, run object detection on live frames while preventing latency buildup and ensuring proper buffer closure.

Quick Start

Use the android-on-device-ml skill to implement a CameraX image analysis loop that runs ML Kit, TensorFlow Lite, or MediaPipe inference and returns structured results in real time on the device.