W3C Publishes Working Draft For Web Neural Network API
The W3C and their stakeholders have been developing the Web Neural Network API as a means of allowing neural network inference hardware acceleration from the browser. This follows the W3C launching a web machine learning working group earlier in the year. Google and Microsoft have been among the companies involved with this working group in looking to be able to exploit dedicated machine learning hardware accelerators and native instructions from within web browsers.
The "WebNN" has been designed with use-cases in mind like person detection, face recognition, super resolution, image captioning, emotion analysis, noise suppression, and more.
For those concerned about arbitrary websites being able to take advantage of machine learning hardware, the security considerations for this new API include making it disabled by default in all cross-origin frames and ties into security found for WebGPU and WebGL.
The first public working draft of the Web Neural Network API can be found at W3.org.
5 Comments
