Human is a machine learning library that provides AI-powered detection and tracking of human physical characteristics and movements, including face detection, body pose, hand tracking, and emotion prediction, running in browsers and Node.js environments.
The library solves the problem of integrating multiple computer vision models into applications by combining face detection, facial analysis, body tracking, hand tracking, iris analysis, and gesture recognition into a single unified API. It works by loading machine learning models on demand, executing them in an attention pipeline where related models run together to provide detailed results only when needed, and applying intelligent temporal interpolation to smooth results across frames regardless of processing speed. The tool includes built-in handling for images, video streams, and webcam input, with optimized preprocessing that can enhance image quality from any source.
Developers should choose this library if they need multiple human-detection capabilities in a single integration rather than assembling separate models. It suits browser-based applications on both desktop and mobile platforms, as well as Node.js server-side processing. The tool is compatible with WebGPU, WebGL, WASM, and CPU backends, and can run in WebWorkers and WebViews. For Node.js, it supports WASM execution on architectures lacking TensorFlow binaries, as well as GPU-accelerated execution via CUDA on supported versions.
The project maintains active release cycles with documented release notes and maintains compatibility across multiple Node.js versions while explicitly noting incompatibility with version 23.x due to breaking changes in TensorFlow dependencies. Development includes comprehensive demo applications covering simple annotated examples and advanced processing interfaces, alongside a gallery of live examples demonstrating the library's capabilities. The codebase is written primarily in HTML with TypeScript support, and the project provides detailed documentation including a wiki of demo applications.
