7 dos and don'ts of using ML on the web with MediaPipe- Google Developers Blog

This post provides seven dos and don'ts for web developers using ML with MediaPipe. Key technical recommendations include: not bundling ML models in the app but storing them in cloud storage and caching them locally, initializing tasks early and warming them up with a placeholder, cleaning up WebAssembly resources using the .close() method, using MediaPipe Studio for testing and customization, testing on different devices, and choosing models based on use case (speed vs. accuracy) rather than defaulting to the largest model. It highlights the use of WebAssembly for running C++ code in the browser and the importance of manual memory management for C++ resources.