Metal GPU Deep Learning and Image processing on the OSX/IOS

Github Repo:

https://github.com/QuantScientist/GPUSwiftKaggler

Machine learning and Image processing on the OSX/IOS GPU

A hands-on workshop with Shlomo Kashani.

Important Note: This is a hands-on workshop using IOS/OSX/Xcode -> Please get a Mac (you can still participate of course without a Mac)

You must also install:

iOS 11.0 Beta

Xcode 9+ Beta

Agenda:

Dreaming about writing an ML based app for the iPhone? Want to get your hands dirty writing basic image processing filters using Swift Plygrounds ... on the GPU using Metal?

You are in the right place.

Apple's Metal is a low-level, low-overhead hardware-accelerated graphics API that debuted in iOS 8. Apple states that Metal "lets developers create highly immersive console-style games on iPhone. Metal is optimized to allow the CPU and GPU to work together to deliver detailed graphics and complex visual effects."

Apple had also recently presented a native machine learning (CoreML) and machine vision (Vision) frameworks augmenting the already existing libraries they provided (e.g, Accelerate). This significantly enhances the vast array of ML based apps that can be deployed to IOS based devices.

In this Hackathon, we will use data set, pre-process it and deploy an app to an iOS target wherein an application will subsequently consume the data and run an ML model using a basic Swift based UI on the GPU using Metal.

Tentative Schedule:

• 17:30 Doors open, get a drink and meet cool data people.

Resources:

https://developer.apple.com/metal/

https://github.com/xmartlabs/Bender

https://github.com/deeplearningkit/deeplearningkit/

https://github.com/mateuszbuda/Sudoku/tree/master/Sudoku

https://github.com/jtbandes/Metalbrot.playground

https://www.invasivecode.com/weblog/metal-image-processing

https://www.invasivecode.com/weblog/convolutional-neural-networks-ios-10-macos-sierra/