MediaPipe-rs is a Rust library for running Google MediaPipe AI inference tasks in the WasmEdge WebAssembly sandbox.
The library solves the problem of building AI inference applications in Rust without Python dependencies by wrapping MediaPipe's task APIs. It supports vision tasks like object detection, image classification, face detection, and hand landmark detection; audio classification; and text classification and embedding. The approach uses a builder pattern where each task type provides a Builder to configure options, a Task instance containing model information, and a Session for executing inference with reusable buffers to minimize allocation overhead.
Developers should choose this tool when they need to run MediaPipe models in Rust environments, particularly within WebAssembly contexts via WasmEdge. It suits projects that require computer vision, audio, or text processing without introducing Python as a dependency. The library supports not only official MediaPipe models but also TensorFlow Hub models and custom TensorFlow Lite models with appropriate metadata. The low-code API design mirrors MediaPipe's Python interface, making it accessible to developers familiar with that ecosystem.
The project maintains active continuous integration checks and publishes regularly to the Rust package registry with accompanying documentation. Development activity shows consistent attention to the codebase with regular updates and maintenance of the build pipeline.
