mil-tokyo/webdnn

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1999 stars · 149 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

WebDNN is a framework for running deep neural network inference directly in the web browser, accepting ONNX models without Python preprocessing. It accelerates execution via WebGPU (WGSL), WebGL2/WebGL1, and WebAssembly backends, with optional offline model optimization.

Use cases

• run onnx models in the browser
• accelerate neural network inference with webgpu
• deploy deep learning models client-side without a server
• run image classification models in a web page
• execute dnn inference with webgl fallback
• optimize onnx models offline for web deployment

When to choose

• you need to run ONNX neural network inference entirely client-side in the browser
• you want GPU acceleration via WebGPU with WebGL and WebAssembly fallbacks
• you want to avoid sending user data to a server for inference
• you need offline model optimization before browser deployment

When to avoid

• you need training or fine-tuning of models, not just inference
• you target server-side or native (non-browser) deployment
• you need broad operator coverage beyond what WebDNN's ONNX support provides
• you require long-term commercial support or a permissive license guarantee (repository license is not standard)

Facets

library · maturity active

machine-learning llm-inference deep-learning gpu-computing deep-learning machine-learning web-development frontend browser onnx webgpu webgl webassembly inference neural-networks typescript web-server nodejs javascript

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mil-tokyo/webdnn")