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
WWDC2025セッション共有: Visionフレームワークによるドキュメントの読み込み
Search
ni_san2000
June 26, 2025
Technology
0
210
WWDC2025セッション共有: Visionフレームワークによるドキュメントの読み込み
in Swift愛好会スピンオフ WWDC25セッション要約会@DeNA (2025/06/26)
ni_san2000
June 26, 2025
Tweet
Share
More Decks by ni_san2000
See All by ni_san2000
Appleの“ホーム”アプリを使いたい! 2024年の国内スマートホーム事情
ryosism
0
160
Other Decks in Technology
See All in Technology
私がよく使うMCPサーバー3選と社内で安全に活用する方法
kintotechdev
0
130
AI時代のIssue駆動開発のススメ
moongift
PRO
0
280
Kiro Meetup #7 Kiro アップデート (2025/12/15〜2026/3/20)
katzueno
2
260
ブラックボックス化したMLシステムのVertex AI移行 / mlops_community_62
visional_engineering_and_design
1
220
How to install a gem
indirect
0
1.8k
OpenClawでPM業務を自動化
knishioka
1
310
パワポ作るマンをMCP Apps化してみた
iwamot
PRO
0
200
VSCode中心だった自分がターミナル沼に入門した話
sanogemaru
0
820
「通るまでRe-run」から卒業!落ちないテストを書く勘所
asumikam
2
820
FastMCP OAuth Proxy with Cognito
hironobuiga
3
220
GitHub Advanced Security × Defender for Cloudで開発とSecOpsのサイロを超える: コードとクラウドをつなぐ、開発プラットフォームのセキュリティ
yuriemori
1
110
OCI技術資料 : ロード・バランサ 概要 - FLB・NLB共通
ocise
4
27k
Featured
See All Featured
Paper Plane
katiecoart
PRO
0
48k
From π to Pie charts
rasagy
0
160
Claude Code のすすめ
schroneko
67
220k
DevOps and Value Stream Thinking: Enabling flow, efficiency and business value
helenjbeal
1
150
The B2B funnel & how to create a winning content strategy
katarinadahlin
PRO
1
310
Side Projects
sachag
455
43k
Fantastic passwords and where to find them - at NoRuKo
philnash
52
3.6k
Typedesign – Prime Four
hannesfritz
42
3k
No one is an island. Learnings from fostering a developers community.
thoeni
21
3.6k
Mozcon NYC 2025: Stop Losing SEO Traffic
samtorres
0
190
Taking LLMs out of the black box: A practical guide to human-in-the-loop distillation
inesmontani
PRO
3
2.1k
Navigating Weather and Climate Data
rabernat
0
150
Transcript
WWDC2025ηογϣϯڞ༗: VisionϑϨʔϜϫʔΫʹΑΔυΩϡϝϯτͷ ಡΈࠐΈ in SwiftѪձεϐϯΦϑ WWDC25ηογϣϯཁձ@DeNA (2025/06/26) ʹʔ͞Μ(@ni_san2000)
ࣗݾհ • ʹʔ͞Μ(@ni_san2000) • ۀͰiOSΞϓϦΛ࡞͍ͬͯ·͢(3) • Apple৴ऀͳͷͰجௐߨԋੲ͔ΒݟͯΔ • macOSͷωʔϛϯάετʔϦʔ •
ΫϨΠάɾϑΣσϦΪͷύϧΫʔϧ • झຯ • Χϝϥ / ؍༿২ / ίʔώʔ / eεϙʔπ • ࠷ۙࣸਅίϯςετͱ͔ڵຯ͋Γ·͢ 2
͢ηογϣϯʹ͍ͭͯ • VisionϑϨʔϜϫʔΫʹΑΔυΩϡϝϯτͷಡΈࠐΈ • “Foundation ModelsͰͳ͍”ɺAIؔ࿈ͷ͓Ͱ͢ 3 ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
VisionϑϨʔϜϫʔΫʹ͍͓ͭͯ͞Β͍ • ಈը૾ʹಛԽͨ͠AIϞσϧͷػೳΛఏڙ͢ΔϑϨʔϜϫʔΫ • ը૾ೝࣝ, ମݕग़, إೝࣝͳͲ • .mlmodelΛཁ͢ΔػೳΛखܰʹAPIͱͯ͠ར༻Ͱ͖Δ •
iOS18Ͱ31ݸͷAPIΛఏڙ͍ͯͨ͠ 4 ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQNBDIJOFMFBSOJOHNPEFMT
VisionϑϨʔϜϫʔΫʹ͍͓ͭͯ͞Β͍ CalculateImageAestheticsScoresRequest ClassifyImageRequest CoreMLRequest DetectAnimalBodyPoseRequest DetectBarcodesRequest DetectContoursRequest DetectDocumentSegmentationRequest DetectFaceCaptureQualityRequest DetectFaceLandmarksRequest
DetectFaceRectanglesRequest DetectHorizonRequest DetectHumanBodyPose3DRequest DetectHumanBodyPoseRequest DetectHumanHandPoseRequest DetectHumanRectanglesRequest DetectRectanglesRequest DetectTextRectanglesRequest DetectTrajectoriesRequest GenerateAttentionBasedSaliencyIma GenerateForegroundInstanceMaskRequest GenerateImageFeaturePrintRequest GenerateOpticalFlowRequest GeneratePersonInstanceMaskRequest GeneratePersonSegmentationRequest RecognizeAnimalsRequest RecognizeTextRequest TrackHomographicImageRegistration TrackObjectRequest TrackOpticalFlowRequest TrackRectangleRequest TrackTranslationalImageRegistrationRequest 5 RecognizeDocumentRequest /FX DetectLensSmudgeRequest /FX
͜Ε·ͰͷυΩϡϝϯτೝࣝ • ݟ͍͑ͯΔจࣈೝࣝͰ͖Δ • ͚ͲɺจষߏදͷίϯςΩετࣦΘΕΔ 6 ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
xOS26͔ΒͰ͖ΔΑ͏ʹͳΔ͜ͱ • υΩϡϝϯτ͔ΒจॻͷߏΛཧղͯ͠ཁૉͷೝ͕ࣝͰ͖Δ • ஈམɾՕॻ͖දͷίϯςΩετೝࣝ • ಛఆͷϑΥʔϚοτʹଈͨ͠ςΩετͷࣝผ • QRίʔυͷऔΓग़͠ 7
υΩϡϝϯτ͔ΒจॻͷߏΛཧղͯ͠ཁૉͷೝ͕ࣝͰ͖Δ ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
ᶃ RequestΫϥεͷ༻ҙ (RecognizeDocumentRequest) ᶄ ը૾Λ͢ ᶅ DocumentObservationΛड͚औΔ ᶆ document͔Β֤ཁૉΛऔಘ ᶇ
That’s it !
DocumentObservationͷݕग़ߏ • ContainerList, Table, Text, BarcodesͳͲͷཁૉΛ࣋ͭ • $FMM*UFNT$POUBJOFSͷཁૉ͍࣋ͬͯΔͷͰɺ͞ΒʹแͰ͖Δ 10 ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
☝
• ࣝผϑΥʔϚοτ10छྨ • ຊޠ·ͩNot supported • දࣔ͢ΔࡍʹText AttributeΛࢦఆ͢Δඞཁ͕ͳ͘ͳΓͦ͏ ςΩετΛࣝผ͢ΔDataDetector 11
આ໌ 5ZQF ิ ΧϨϯμʔ CalendarEvent ϝʔϧΞυϨε EmailAddress ϑϥΠτ൪߸ FlightNumber 63- Link ଌఆ ୯Ґ͖ Measurement %JNFOTJPOͰදݱͰ͖Δ୯Ґ ֹۚ ௨՟͖ MoneyAmount -PDBMF$VSSFODZͰදݱͰ͖Δ୯Ґ ࢧ͍ঢ়گ PaymentIdentifier 6OJ fi FE1BZNFOUT*OUFSGBDF 61* ి൪߸ PhoneNumber ॅॴ PostalAddress FHTUSFFU DJUZ TUBUF ૹ൪߸ ShipmentTrackingNumber ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
֤ϑΥʔϚοτͱͯ͠σʔλΛऔಘ • switch-case͢Δ͚ͩ • ͲͷϑΥʔϚοτ͔ςΩετ୯ҐͰࣝผࡁΈ 12
υΩϡϝϯτ͔ΒจॻͷߏΛཧղͯ͠ཁૉͷೝ͕ࣝͰ͖Δ ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
• VisionϑϨʔϜϫʔΫͷ৽نAPI • υΩϡϝϯτͷ༰ΛߏతʹੳͰ͖Δ • ContainerTable, List, Cell, TextͳͲͷཁૉΛ͍࣋ͬͯΔ •
CellListͷItemೖΕࢠͰContainerΛ֨ೲͰ͖Δ • Text֨ೲ͞Εͨจࣈ͔ΒϑΥʔϚοτΛೝࣝͰ͖Δ • DataDetector.Match.SemanticDetails.XXXXXʹଟͷ৽نσʔλߏ RecognizeDocumentRequest·ͱΊ 14
ΓͷVisionؔ࿈ͷ৽ཁૉ̎ͭ ͓·͚ 15
• DetectLensSmudgeRequest / SmudgeObservation͕Ճ • ೖྗͨ͠ը૾͕ԚΕ͍ͯΔ͔Λผ͢Δ • ᮢΛઃ͚Δ͜ͱͰੳΤϥʔΛະવʹ͛Δ Ϩϯζද໘ͷԚΕಶΓΛݕग़͢ΔObservation 16
ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
• खͷؔઅҐஔΛݕग़͢ΔϞσϧ͕ߋ৽͞Εͨ • ਪͷਫ਼ͱ্͕ • WWDC2021Ͱൃද͞ΕͨϞσϧ͔ΒΞοϓσʔτ HandPose DetectionͷੳϞσϧߋ৽ 17 ႄႨჭğIUUQTEFWFMPQFSBQQMFDPNKQWJEFPTQMBZXXED
• VisionϑϨʔϜϫʔΫͷػೳՃ • υΩϡϝϯτͷ༰ΛߏతʹੳͰ͖Δ • ςΩετͷ༰ΛྨͰ͖Δ • ͦͷଞʹ2ͭͷVisionϑϨʔϜϫʔΫͷΞοϓσʔτ • DetectLensSmudgeRequest
/ SmudgeObservation • HandPose DetectionͷੳϞσϧߋ৽ ͓͠·͍ 18