Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
CoreMLではじめる機械学習
Search
naru-jpn
June 21, 2017
Technology
0
1.1k
CoreMLではじめる機械学習
Neural Networks on Keras ( TensorFlow backends )
naru-jpn
June 21, 2017
Tweet
Share
More Decks by naru-jpn
See All by naru-jpn
配信アプリのためのリアルタイムプッシュ通知ぼかしの夢
narujpn
3
870
PiPを応用した配信コメントバー機能の開発秘話と技術の詳解 / pip_streaming_comment_bar
narujpn
3
4k
Updating an App to Use Swift Concurrency 解説
narujpn
2
320
PiP で実現するミラティブの配信コメントバー / pip-streaming-comment-bar
narujpn
0
1.1k
App Extension のスタックトレース情報からクラッシュを解析/集計する / Analyzing app extension's stack trace
narujpn
3
1.4k
ミラティブとWebRTC - WebRTC framework の中身を覗いてみよう / WebRTC framework AudioUnit Processing
narujpn
1
2.1k
CoreML3のオンデバイストレーニングでつくる母音推定
narujpn
0
420
AltConfと周辺の歩き方
narujpn
0
2k
エンジニア経験を活かしたスクラムマスターとして 開発チームとプロダクトを成長させる
narujpn
1
400
Other Decks in Technology
See All in Technology
Winning at PHP in Production in 2025
beberlei
1
230
【Oracle Cloud ウェビナー】ご希望のクラウドでOracle Databaseを実行〜マルチクラウド・ソリューション徹底解説〜
oracle4engineer
PRO
1
130
2025-04-24 "Manga AI Understanding & Localization" Furukawa Arata (CyberAgent, Inc)
ornew
2
300
Microsoft の SSE の現在地
skmkzyk
0
230
Oracle Cloud Infrastructure:2025年4月度サービス・アップデート
oracle4engineer
PRO
0
230
Microsoft Fabric vs Databricks vs (Snowflake) -若手エンジニアがそれぞれの強みと違いを比較してみた- "A Young Engineer's Comparison of Their Strengths and Differences"
reireireijinjin6
1
110
AIにおけるソフトウェアテスト_ver1.00
fumisuke
1
300
Terraform Cloudで始めるおひとりさまOrganizationsのすゝめ
handy
2
210
OPENLOGI Company Profile
hr01
0
63k
Рекомендации с нуля: как мы в Lamoda превратили главную страницу в ключевую точку входа для персонализированного шоппинга. Данил Комаров, Data Scientist, Lamoda Tech
lamodatech
0
830
Mastraに入門してみた ~AWS CDKを添えて~
tsukuboshi
0
360
OPENLOGI Company Profile for engineer
hr01
1
25k
Featured
See All Featured
Templates, Plugins, & Blocks: Oh My! Creating the theme that thinks of everything
marktimemedia
30
2.3k
Practical Orchestrator
shlominoach
187
11k
YesSQL, Process and Tooling at Scale
rocio
172
14k
ピンチをチャンスに:未来をつくるプロダクトロードマップ #pmconf2020
aki_iinuma
119
51k
The Power of CSS Pseudo Elements
geoffreycrofte
75
5.8k
Easily Structure & Communicate Ideas using Wireframe
afnizarnur
194
16k
Into the Great Unknown - MozCon
thekraken
38
1.7k
Why Our Code Smells
bkeepers
PRO
336
57k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
29
1.6k
Art, The Web, and Tiny UX
lynnandtonic
298
20k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
136
33k
Transcript
CoreMLͰ͡ΊΔػցֶश Neural Networks on Keras ( TensorFlow backends ) Timers
inc. / Github: naru-jpn / Twitter: @naruchigi
CoreMLͰ͡ΊΔػցֶश Timers inc. / Github: naru-jpn / Twitter: @naruchigi Neural
Networks on Keras ( TensorFlow backends )
What is Neural Networks?
One of machine learning models. - Neural networks - Tree
ensembles - Support vector machines - Generalized linear models - … https://developer.apple.com/documentation/coreml/converting_trained_models_to_core_ml
What is Keras?
Theano TensorFlow Keras Keras is a high-level neural networks API,
written in Python and capable of running on top of either TensorFlow, CNTK or Theano. https://keras.io
What is CoreML?
Accelerate and BNNS Metal Performance Shaders CoreML BNNS : Basic
Neural Network Subroutines https://developer.apple.com/documentation/coreml With Core ML, you can integrate trained machine learning models into your app. Core ML requires the Core ML model format.
CoreML Trained Model Application Keras Train coremltools
What is coremltools?
Convert existing models to .mlmodel format from popular machine learning
tools including Keras, Caffe, scikit-learn, libsvm, and XGBoost. https://pypi.python.org/pypi/coremltools coremltools
CoreML Trained Model Application Keras Train coremltools
(Demo App)
Environment - Tensorflow 1.1.0 (virtualenv) - Keras 1.2.2 - coremltools
0.3.0 - Xcode 9.0 beta ※ Tensorflow, Keras coremltools ͷରԠόʔδϣϯͰ͋Δඞཁ͕͋ΔͷͰগ͠ݹ͍Ͱ͢ɻ
Programs to train neural networks - mnist_mlp.py - mnist_cnn.py ※
Keras ͷ࠷৽όʔδϣϯͷϦϯΫʹͳ͍ͬͯ·͕͢ɺ࣮ࡍόʔδϣϯ 1.2.2 Λࢀর͠·͢ɻ https://github.com/fchollet/keras/tree/master/examples
Convert model with coremltools 1. Import coremltools import coremltools model
= Sequential() … coreml_model = coremltools.converters.keras.convert(model) coreml_model.save("keras_mnist_mlp.mlmodel") 2. Convert model
Import model into Xcode project // 入力データ class keras_mnist_mlpInput :
MLFeatureProvider { var input1: MLMultiArray // … } // 出力データ class keras_mnist_mlpOutput : MLFeatureProvider { var output1: MLMultiArray // … } // モデル @objc class keras_mnist_mlp:NSObject { var model: MLModel init(contentsOf url: URL) throws { self.model = try MLModel(contentsOf: url) } // … func prediction(input: keras_mnist_mlpInput) throws -> keras_mnist_mlpOutput { // … keras_mnist_mlp.mlmodel Λѻ͏ҝͷίʔυ͕ࣗಈੜ͞ΕΔ
Prepare model and input in code // モデルの作成 let model
= keras_mnist_mlp() // 入力データの格納用変数 (入力は28*28の画像) let input = keras_mnist_mlpInput( input1: try! MLMultiArray(shape: [784], dataType: .double) )
Modify input value // 入力データの 0 番目の要素に 1.0 を代入 input.input1[0]
= NSNumber(value: 1.0)
Make a prediction // モデルに入力データを渡して計算 let output = try model.prediction(
input: self.input )
CoreML Trained Model Application Keras Train coremltools Recap
Demo App on Github https://github.com/naru-jpn/MLModelSample
͝ਗ਼ௌ͋Γ͕ͱ͏͍͟͝·ͨ͠