Each inference takes about 110 ms in chrome on a core i3. And memory consumption is ~ about a tab with Facebook open.

Of course a native client would be much efficient considering Tensorflow.js runs over WebGL on top of browser.

What do you suggest for the name?