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 First Impressions
Search
Chiharu Nameki
June 19, 2017
Technology
880
1
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Core ML First Impressions
CA.swift #3 WWDC17報告会
Chiharu Nameki
June 19, 2017
More Decks by Chiharu Nameki
See All by Chiharu Nameki
iPhoneでの空間メディア視聴体験の探求
ridwy
1
83
空間再現力の鍵、APMPを読み解く
ridwy
1
430
空間ビデオフォーマット、MV-HEVCが写し撮る世界
ridwy
4
2.8k
空間オーディオを活用しよう!
ridwy
1
710
AVPlayerのHLS再生時のエラーをハンドリングする
ridwy
0
910
AirPlayの裏舞台
ridwy
0
3k
AVPlayerできちんとコンテンツ保護
ridwy
0
2.2k
検証 SharePlay media playback
ridwy
1
2.1k
カスタムInstrumentsでアプリをピカピカに!
ridwy
5
13k
Other Decks in Technology
See All in Technology
[2026-07-15] AI Ready なはずだったアーキテクチャと、見えてきた課題・次に目指す状態
wxyzzz
4
2.7k
Empower GenAI with Agile - あなたのアジャイルが生成AIのバフになる仕組み
hageyahhoo
1
180
Amazon EVS で VCF 9.0 / 9.1 のサポート開始まとめ
mtoyoda
0
290
ボーイスカウトルールでメモリやスキルを改善しよう
azukiazusa1
1
700
型は壁、Rustでもバグを直すな、表現できなくせよ
nwiizo
13
2k
事業価値を⽣み出すSREへ SREが担うべき意思決定の5層
kenta_hi
2
3.5k
LLMやAIエージェントをソフトウェアに組み込むプラクティス
shibuiwilliam
1
360
10年目を迎えた「ABEMA」がどのように AI 活用を推進して、AI 駆動開発にシフトしているのか / How ABEMA, entering its 10th year, is promoting the use of AI and shifting toward AI-driven development
miyukki
0
120
ループエンジニアリングでE2Eテストを実践
noriyukitakei
0
350
SRE Lounge Hiroshimaへの招待
grimoh
0
640
AIと共生する開発者プラットフォーム:バクラクのモノレポ×マイクロサービス基盤
sakajunquality
2
3.4k
穢れた技術選定について
watany
3
310
Featured
See All Featured
Statistics for Hackers
jakevdp
799
230k
Art, The Web, and Tiny UX
lynnandtonic
304
22k
Darren the Foodie - Storyboard
khoart
PRO
3
3.4k
30 Presentation Tips
portentint
PRO
1
350
Agile Leadership in an Agile Organization
kimpetersen
PRO
0
190
Six Lessons from altMBA
skipperchong
29
4.3k
Mobile First: as difficult as doing things right
swwweet
225
10k
Into the Great Unknown - MozCon
thekraken
41
2.6k
Organizational Design Perspectives: An Ontology of Organizational Design Elements
kimpetersen
PRO
1
760
Building a A Zero-Code AI SEO Workflow
portentint
PRO
0
630
The SEO Collaboration Effect
kristinabergwall1
1
500
Design in an AI World
tapps
1
260
Transcript
Core ML First Impressions CA.swift #3 WWDC17 ใࠂձ 2017/6/19 ߦ
ઍय़
ࣗݾհ • ߦ ઍय़ʢͳΊ͖ ͪΔʣ • AmebaΞϓϦ iOSΤϯδχΞ • 20153݄த్ೖࣾ
macOS͚ΞϓϦϛυϧΣΞΛ࡞ͬͯ·ͨ͠ • WWDCͷࢀՃ2015Ҏདྷ̎ ←!?
Core ML Framework
WWDC17 Keynote
• ΫϥΠΞϯταΠυͰػցֶशͷ݁ՌͷϞσϧΛར༻͢ΔͨΊͷI/F ͜Ε·ͰެࣜͷࣸਅΞϓϦͳͲͰར༻͞Ε͍ͯͨػೳ͕3rd Partyʹղ์͞Εͨɻ • ϋʔυΣΞʹΑΔߴͳॲཧɻCPU, GPUΛࣗಈతʹબɻ • ্Ͱ࣮ߦ͢ΔͨΊҎԼͷϝϦοτ͕͋Δɻ
Underlying Function Introducing Core ML
Accelerate & BNNS Metal Performance Shaders CPU GPU Core ML
Vision Natural Language Processing APP NEW NEW Architecture
Accelerate & BNNS Metal Performance Shaders CPU GPU Core ML
Vision NLP NSLinguisticTagger APP iOS 4.0+, macOS 10.3+, tv0S 9.0+, watchOS 4.0+ iOS 11.0+, macOS 10.3+, tv0S 11.0+, watchOS 4.0+ iOS 9.0+, macOS 10.13+, tv0S 9.0+ iOS 11.0+, macOS 10.13+, tv0S 11.0+ iOS 5.0+, macOS 10.7+, tv0S 9.0+, watchOS 2.0+ OS version & Hardware limitations Metal supported A7ʢ= iPhone 5sʣand later NEW NEW
Platforms State of the Union ։ൃϑϩʔ
Core ML in Depth ։ൃϑϩʔ
https://developer.apple.com/documentation/ coreml/converting_trained_models_to_core_ml Converting Trained Models to Core ML ΑΓ •
CaffeͳͲͰੜͨ͠Ϟσϧ Λ .mlmodel ϑΝΠϧʹมͯ͠ར༻ • มπʔϧ coremltool ʢOpen Sourceʣ • coremltoolΛࢀߟʹมπʔϧͷ࡞ Մೳ αϙʔτπʔϧ & Ϟσϧ
API Core ML Framework MLModel MLFeatureProvider MLFeatureProvider INPUT OUTPUT func
prediction(from input: MLFeatureProvider) throws -> MLFeatureProvider
API Core ML Framework MLFeatureProvider featureNames featureValue(for: String) MLFeatureValue type
int64Value doubleValue stringValue dictionaryValue: [AnyHashable: NSNumber] imageBufferValue: CVPixelBuffer? multiArrayValue: MLMultiArray? ಛྔͷϥούʔ ಛྔʹΞΫηε͢ΔͨΊͷI/F ͷଟ࣍ݩྻ
.mlmodel͔Βࣗಈੜ͞ΕΔίʔυ MLModel MLFeatureProvider MLFeatureProvider ModelInput ModelOutput func prediction(input: ModelInput) throws
-> ModelOutput Model ※ Model࣮ࡍʹϞσϧ໊ʹͳΔ
Vision Framework
Vision Framework • Core ML frameworkͷ্ʹΓཱͭɺComputer Visionͷ ྖҬʹಛԽͨ͠API • ඇৗʹଟ͘ͷೖྗϑΥʔϚοτΛαϙʔτ
• CVPixelBuffer, CGImage, CIImage, NSURL, NSData • CVPixelBufferʹมͯ͠Core MLʹͯ͘͠ΕΔ • ը૾ͷΦϦΤϯςʔγϣϯղੳʹదͨ͠αΠζͷεέʔ Ϧϯάαϙʔτ iOS 11.0+, macOS 10.13+, tv0S 11.0+
Vision Framework • إݕग़ • إύʔπݕग़ • ը૾߹ʢύϊϥϚͳͲʣ • ۣܗݕग़
• όʔίʔυݕग़ • ςΩετݕग़ • τϥοΩϯά • ਫฏݕग़ • MLModelΛར༻ͨ͠ղੳ
Vision Framework Deep LearningΛར༻͍ͯ͠ΔͷͰ ԣإɾإͷҰ෦͕ӅΕ͍ͯΔɾࢠɺαϯάϥεண༻࣌ͳͲOK
API Vision Framework VNRequest VNRequestHandler VNObservation completionHandler handler.perform([request]) ೖྗը૾ &
ϦΫΤετͷॲཧ ϦΫΤετ ݁Ռ VNDetectFaceRectanglesRequest VNDetectFaceLandmarksRequest VNDetectRectanglesRequest VNCoreMLRequest VNDetectBarcodesRequest VNTrackingRequest VNDetectHorizonRequest … VNImageRequestHandler VNSequenceRequestHandler VNFaceObservation VNTextObservation VNHorizonObservation VNRectangleObservation VNBarcodeObservation VNTrackObjectRequest VNClassificationObservation
Demo
Core ML First Impressions • ϞσϧͷऔΓࠐΈඇৗʹ؆୯ • ϓϩμΫτʹऔΓࠐΉͷਫ਼ݹ͍σόΠεͰͷ Λݟͳ͕Β •
ೖྗͷิॿͳͲʹྑͦ͞͏ • Ϟσϧͷ՝Λղܾ͢Δඞཁ͕͋Δ • ༰ྔΛখ͘͞ • ਫ਼্
ࢀߟηογϣϯ • WWDC17 Keynote • Platforms State of the Union
• Introducing Core ML • Core ML in Depth • Vision Framework: Building on Core ML • Natural Language Processing and your Apps
Thank you. Apple Park June 4th 2017. Still Constructing …