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
Core ML / Vision Frameworkを使ってできること / What can ...
Search
Shinichi Goto
June 30, 2017
Programming
3
4.3k
Core ML / Vision Frameworkを使ってできること / What can we achieve using Core ML and Vision framework
2017/06/30 WWDC - Developer's Living #lifull_wwdc
Shinichi Goto
June 30, 2017
Tweet
Share
More Decks by Shinichi Goto
See All by Shinichi Goto
WWDC18 ML Overview
_shingt
1
1.4k
Core ML 🏃 iOS Engineer
_shingt
1
590
Wantedly Peopleのスキャン画面の裏側 / Wantedly People Scanning Screen
_shingt
6
6.4k
Providing Better Feedback in Real-time Object Detection Apps
_shingt
2
1.5k
Value Types in WWDC16
_shingt
3
2.6k
debug-remote-local-notification-on-watchos
_shingt
0
11k
Server Side Swift question
_shingt
3
1.1k
Other Decks in Programming
See All in Programming
GitHub Actions × AWS OIDC連携の仕組みと経緯を理解する
ota1022
0
240
Model Pollution
hschwentner
1
180
WebエンジニアがSwiftをブラウザで動かすプレイグラウンドを作ってみた
ohmori_yusuke
0
170
Serena MCPのすすめ
wadakatu
4
900
複雑化したリポジトリをなんとかした話 pipenvからuvによるモノレポ構成への移行
satoshi256kbyte
1
780
iOSエンジニア向けの英語学習アプリを作る!
yukawashouhei
0
180
デミカツ切り抜きで面倒くさいことはPythonにやらせよう
aokswork3
0
190
ててべんす独演会〜Flowの全てを語ります〜
tbsten
1
220
CI_CD「健康診断」のススメ。現場でのボトルネック特定から、健康診断を通じた組織的な改善手法
teamlab
PRO
0
180
Conquering Massive Traffic Spikes in Ruby Applications with Pitchfork
riseshia
0
150
ソフトウェア設計の実践的な考え方
masuda220
PRO
3
490
Breaking Up with Big ViewModels — Without Breaking Your Architecture (droidcon Berlin 2025)
steliosf
PRO
1
330
Featured
See All Featured
Dealing with People You Can't Stand - Big Design 2015
cassininazir
367
27k
Music & Morning Musume
bryan
46
6.8k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
140
34k
Six Lessons from altMBA
skipperchong
28
4k
Unsuck your backbone
ammeep
671
58k
How STYLIGHT went responsive
nonsquared
100
5.8k
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
229
22k
How to Think Like a Performance Engineer
csswizardry
27
2k
Why Our Code Smells
bkeepers
PRO
339
57k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
32
2.2k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
48
9.7k
Designing for humans not robots
tammielis
254
25k
Transcript
Core ML / Vision Framework ΛͬͯͰ͖Δ͜ͱ ɹ 2017/06/30 WWDC -
Developer's Living @ LIFULL shingt (Shinichi Goto)
shingt (Shinichi Goto) GitHub: @shingt Twi5er: @_shingt 2
Core ML Vision Framework 3
4
Outline • Core MLͷ֓ཁ • Vision Frameworkͷ֓ཁ • Ͱ͖Δ͜ͱ /
ࣄྫհ 5
Core ML 6
ML (Machine Learning) 7
8
9
10
Core ML • ֶशࡁͷModelΛར༻ͯ͠ͷਪʹಛԽ • Core ML model format (**.mlmodel)
• Xcode͕Swi6ͷΠϯλʔϑΣΠεΛࣗಈੜ • αϯϓϧϞσϧApple͕ެ։ • Accerelate / Metal্ʹࡌ͓ͬͯΓϋΠύϑΥʔϚϯε • coremltools 11
ɹ let animalModel = AnimalModel() if let prediction = try?
animalModel.prediction(animalImage: image) { return prediction.animalType } 12
ɹ let animalModel = AnimalModel() if let prediction = try?
animalModel.prediction(animalImage: image) { return prediction.animalType } 13
ɹ let animalModel = AnimalModel() if let prediction = try?
animalModel.prediction(animalImage: image) { return prediction.animalType } 14
ɹ let animalModel = AnimalModel() if let prediction = try?
animalModel.prediction(animalImage: image) { return prediction.animalType } 15
coremltools • "iOS্Ͱѻ͏ͨΊͷModelΛͲ͏༻ҙ͢Δ͔" ͷղܾࡦ • ओཁͳػցֶशπʔϧͷֶशࡁModelΛCore ML༻ͷModelม • Keras,
Caffe, scikit-learn, etc. 16
17
6/28ʹKeras 2.0αϙʔτʢൃද࣌1.2.2ͷΈͩͬͨʣ h"ps:/ /forums.developer.apple.com/thread/81196 18
Vision Framework 19
Vision Framework • Core ML্ʹࡌͬͨը૾ೝࣝɾମݕग़ͳͲͷը૾ղੳ༻ͷϑϨʔϜϫʔΫ • Detec,on • Face, Face
landmarks, Rectangle, Barcode, Text, Horizon • طଘͷͷਫ਼্ʢDeep Learningͷ׆༻ʣ • Tracking • Image Registra,on • Core MLͱͷΈ߹Θͤ 20
21
Tracking • ը૾ʢಈըʣதͷମͷ • إͷTrackingCIDetectorͰՄೳͩͬͨ • ҙͷରʹରͯ͠ͷTracking͕Մೳʹ • VisionͰͷݕग़݁Ռ •
ҙͷྖҬࢦఆ 22
23
Demo (Rectangle Detec,on + Tracking) h"ps:/ /github.com/shingt/VisionTrackerSample 24
զʑCV/MLͷΤΩεύʔτͰ͋Δඞཁͳ͍ ʢͱɺAppleηογϣϯதʹݴ͍ͬͯΔʣ 25
Կ͕Ͱ͖Δͷ͔ʁ ʢΞϓϦέʔγϣϯΤϯδχΞͱͯ͠ͷࢹ͔Βʣ 26
27
28
ࣄྫհ 29
ମݕग़ 30
31
YOLO • YOLO (You only look once) • ߴͳ͜ͱ͕ಛͷମݕग़༻ͷ χϡʔϥϧωοτϫʔΫ
• h1ps:/ /www.youtube.com/watch? v=VOC3huqHrss • ͜ΕҰൠతͳYOLO 32
• iOSࣄྫ • YOLO: Core ML versus MPSNNGraph • Core
MLΛ༻͍ͯiOS্ͰYOLOΛಈ࡞ • Tiny YOLOʢެ։͞Ε͍ͯΔModelʣΛར༻ 33
34
ը૾ੜ 35
Goodfellow, Ian J.; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley,
David; Ozair, Sherjil; Courville, Aaron; Bengio, Yoshua. GeneraIve Adversarial Networks. arXiv:1406.2661, 2014. 36
Alec Radford, Luke Metz, and Soumith Chintala. Unsupervised representa>on learning
with deep convolu>onal genera>ve adversarial networks. arXiv preprint arXiv:1511.06434, 2015. 37
• GAN (Genera+ve Adversarial Nets) • ֶशσʔλͱࣅͨσʔλΛੜ͢ΔϞσϧͷҰछ • iOSࣄྫ •
Crea+ve AI on the iPhone: Genera+ve Adversarial Networks (GAN) with Apple's CoreML Tools • MNISTΛσʔληοτͱͯ͠ɺCore MLΛ༻͍ͯiOS্Ͱࣈ ʢʹࣅͨʣը૾Λੜ 38
39
Summary • Core ML / Vision Framework • iOS্Ͱͷը૾ղੳٕज़ͷར༻ϋʔυϧ͕Լ •
ͱ͍͑ࣝ͋Δఔඞཁʢͱײͨ͡ʣ • Ͱ͖Δ͜ͱ • ը૾ೝࣝ / τϥοΩϯά / ମݕग़ / ը૾ੜ / etc. • Follow @mhollemans 40
ࢀߟηογϣϯ • Introducing Core ML • Core ML in depth
• Vision Framework: Building on Core ML 41
ࢀߟࢿྉ • iOS 11: Machine Learning for everyone • Google’s
MobileNets on the iPhone • YOLO: Core ML versus MPSNNGraph • CreaAve AI on the iPhone: GeneraAve Adversarial Networks (GAN) with Apple's CoreML Tools - Zedge • Why Core ML will not work for your app (most likely) • θϩ͔Β࡞ΔDeep Learning 42
Thanks! 43
ʢิʣͰ͖ͳ͍͜ͱ / ੍ͳͲ • ֶशෆՄ • αϙʔτ͍ͯ͠ΔػցֶशϑϨʔϜϫʔΫʹ͍ͭͯɺಛఆͷόʔδϣϯʹറΒΕΔʢগͳ͘ͱ ݱঢ়ʣ • Kerasͷ2.0αϙʔτೖͬͨ͠ɺࠓޙ͍͛ͯ͘ͷ͔
• ModelͷαΠζ͕େ͖͗͢Δ • RegressionͱClassifica5onͷΈʢ☓ ΫϥελϦϯάɺϥϯΩϯάֶशɺetc.ʣ • ϥϯλΠϜͰϢʔβͷೖྗɾߦಈΛModelʹөͤ͞Δ͜ͱͰ͖ͳ͍ • etc. 44