ML & JSNo you’re not dreaming
View Slide
@yashints
@yashintsOr is it?➔
Lead consultant @ReadifyMeet Yaser@yashintsyashints.dev
Why do it in browser?Wide distributionInteractiveSensorsData stays on the client
@yashintshttps://playground.tensorflow.org
@yashintsTensorflow.js
@yashintsAllows inference andtraining entirely in thebrowser or Node.jsReleased March 2018 GPU-accelerated viaWebGLTensorFlow.js
@yashintsEmpower a diverse group of developersEnable ML peeps to port their work to web
@yashintsBuild your ownRun existingmodelsRetrainwith transfer learning
@yashintsTF TPUTF GPUCore APIWebGLLayers APIModelBrowser Node.jsTF CPUKeras ModelTensorflowSaved models
@yashintsPre-trained modelshttps://github.com/tensorflow/tfjs-models• MobileNet Object detection• PoseNet Human pose detection• CocoSSD Object localisation• BodyPix Human segmentation• USE Text classification
@yashintsBrowser examples
@yashintsBodyPixdemo
@yashintsCreatabilityyas.fyi/2Zl4N2p
@yashintsAirbnb
@yashintsSpeech command
@yashintsNode.js & Desktop
@yashintsyas.fyi/2GBJF0sDoctor Clinic
@yashintsyas.fyi/2LUfBkXMagenta Studio
@yashintsRetraining in Node.js
@yashintsDemo
@yashints“A year spent in artificialintelligence is enough to make onebelieve in God.”“Alan Perlis
@yashintsReferenceshttps://www.tensorflow.org/jshttps://github.com/tensorflow/tfjs-exampleshttps://www.freecodecamp.org/news/tensorflowjs-full-course/https://codelabs.developers.google.com/https://yashints.dev/blog/2018/11/27/get-started-with-tensorflowjs
Reach outThanks!@yashintsyashints.dev