{
  "format": "agent-article-v0",
  "article_id": "EA-072",
  "canonical_url": "/tools/ea-072-vision-video-ai",
  "page_type": "reference",
  "claim_ids": [],
  "evidence_ids": [
    "EVD-EA-072-URL-599FF2E7CFF3",
    "EVD-EA-072-URL-013C0C535D25",
    "EVD-EA-072-URL-BBEA16635537",
    "EVD-EA-072-URL-FFB980AE9E3E",
    "EVD-EA-072-URL-7E5709CA7B06",
    "EVD-EA-072-URL-FD0A9341E4D8",
    "EA-072-R11-RESEARCH",
    "EA-072-R11-WEB-READBACK"
  ],
  "unknown_ids": [
    "EA-072-UNKNOWN-01",
    "EA-072-UNKNOWN-02",
    "EA-072-UNKNOWN-CURRENTNESS"
  ],
  "operational_object_ids": [],
  "provider": [],
  "models": [],
  "plans": [],
  "surfaces": [],
  "regions": [],
  "current_status": "UNKNOWN",
  "observation_status": "SPEC_ONLY",
  "freshness": "UNKNOWN",
  "last_verified": null,
  "supersedes": [],
  "superseded_by": [],
  "provenance": [
    "COMM-068-EA0072",
    "SRC-EA-072-URL-599FF2E7CFF3",
    "SRC-EA-072-URL-013C0C535D25",
    "SRC-EA-072-URL-BBEA16635537",
    "SRC-EA-072-URL-FFB980AE9E3E",
    "SRC-EA-072-URL-7E5709CA7B06",
    "SRC-EA-072-URL-FD0A9341E4D8",
    "WEBBODY-EA0072-R8-C765F7BA"
  ],
  "article_sources": [
    {
      "id": "EVD-EA-072-URL-599FF2E7CFF3",
      "evidence_class": "RESEARCH",
      "title": "official product spec | “What is Image Analysis?” | Microsoft |",
      "href": "https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/overview-image-analysis",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-072-URL-599FF2E7CFF3"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "- official product spec | “What is Image Analysis?” | Microsoft | https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/overview-image-analysis"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-072-URL-013C0C535D25",
      "evidence_class": "RESEARCH",
      "title": "official product overview | “Vision AI: Image and visual AI tools” | Google Cloud |",
      "href": "https://cloud.google.com/vision",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-072-URL-013C0C535D25"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "- official product overview | “Vision AI: Image and visual AI tools” | Google Cloud | https://cloud.google.com/vision"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-072-URL-BBEA16635537",
      "evidence_class": "RESEARCH",
      "title": "official product spec | “Detecting objects and concepts” | Amazon Web Services |",
      "href": "https://docs.aws.amazon.com/rekognition/latest/dg/labels.html",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-072-URL-BBEA16635537"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "- official product spec | “Detecting objects and concepts” | Amazon Web Services | https://docs.aws.amazon.com/rekognition/latest/dg/labels.html"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-072-URL-FFB980AE9E3E",
      "evidence_class": "RESEARCH",
      "title": "official API/spec | “Using the image and video moderation APIs” | Amazon Web Services |",
      "href": "https://docs.aws.amazon.com/rekognition/latest/dg/moderation-api.html",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-072-URL-FFB980AE9E3E"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "- official API/spec | “Using the image and video moderation APIs” | Amazon Web Services | https://docs.aws.amazon.com/rekognition/latest/dg/moderation-api.html"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-072-URL-7E5709CA7B06",
      "evidence_class": "RESEARCH",
      "title": "research benchmark | “Visual Reasoning Benchmark: Evaluating Multimodal LLMs on Classroom-Authentic Visual Problems from Primary Education” | Huti et al. |",
      "href": "https://arxiv.org/abs/2602.12196",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-072-URL-7E5709CA7B06"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "- research benchmark | “Visual Reasoning Benchmark: Evaluating Multimodal LLMs on Classroom-Authentic Visual Problems from Primary Education” | Huti et al. | https://arxiv.org/abs/2602.12196"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-072-URL-FD0A9341E4D8",
      "evidence_class": "RESEARCH",
      "title": "research benchmark | “VisR-Bench: An Empirical Study on Visual Retrieval-Augmented Generation for Multilingual Long Document Understanding” | Chen et al. |",
      "href": "https://arxiv.org/abs/2508.07493",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-072-URL-FD0A9341E4D8"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "- research benchmark | “VisR-Bench: An Empirical Study on Visual Retrieval-Augmented Generation for Multilingual Long Document Understanding” | Chen et al. | https://arxiv.org/abs/2508.07493"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EA-072-R11-RESEARCH",
      "evidence_class": "RESEARCH",
      "title": "EA-072 監査済みresearch / canonical revision binding",
      "supports": [
        "A. Vision capabilityを少なくとも (a) OCR/text reading, (b) image classification/tagging, (c) object detection/localization, (d) caption/dense description, (e) visual question answering/reasoning, (f) multimodal embedding/retrieval, (g) moderation/safety classification, (h) video temporal analysis に分離する。単一“vision capable” flagは禁止候補。\nB. 専用Vision APIはfeature taxonomy、confidence、bounding box/timestamp、model version等のstructured outputを持ち得る。一方MLLMの自然言語回答は同じ検出保証ではない。専用detectorのconfidenceとLLMの自己申告confidenceを同一尺度にしない。\nC. version driftは実在する。Azure Image Analysisはv4.0/v3.2でfeature setが異なり、一部featureはretire済み。current product claimはAPI/model revisionを固定する。\nD. moderation labelはpolicy decisionそのものではない。AWSもconfidence thresholdとtaxonomyを返し、application側が適否を決める構造。model output→policy mapping→human/escalationを分離する。\nE. visual benchmarkは能力のfailure classを示すが、商用tenant/modelの保証にはならない。VRBのspatial reasoning weakness、VisR-Benchのtable/low-resource language weaknessはTEST_PROFILE設計根拠に限定する。"
      ],
      "limit": "canonical監査・revision bindingの存在を示す。外部ソースcurrentnessが本文またはcontrol-planeで確定していない場合は、currentnessをUNKNOWNのまま保持する。",
      "origin_class": "AUDITED_CANONICAL_RESEARCH_PACKET",
      "source_or_experiment_ids": [
        "COMM-068-EA0072"
      ],
      "supporting_locators": [
        "COMM-068-EA0072",
        "EA-072-R5-LEGACY-MIGRATION-20261008-R1",
        "content/staging/EA-072.md"
      ],
      "scope": "監査済みEA本文・canonical revision・既存Source Graph relationの範囲。欠損currentnessは推測で閉じない。",
      "provenance_status": "AUDIT_PASS_CANONICAL_BINDING"
    },
    {
      "id": "EA-072-R11-WEB-READBACK",
      "evidence_class": "IMPL",
      "title": "EA-072 Web body identity / canonical readback",
      "supports": [
        "canonical staging bodyのSHA-256一致とR9 canonical ledger readbackをProduction projectionの入力として検証する。"
      ],
      "limit": "body identity/readbackは本文claim自体の真偽や外部ソースの最新性を追加保証しない。",
      "checked_at": null,
      "origin_class": "IMPLEMENTATION_READBACK",
      "source_or_experiment_ids": [
        "WEBBODY-EA0072-R8-C765F7BA"
      ],
      "supporting_locators": [
        "WEBBODY-EA0072-R8-C765F7BA",
        "content/staging/EA-072.md",
        "EA-072-R5-LEGACY-MIGRATION-20261008-R1"
      ],
      "scope": "R9/R10 canonical readback後のProduction生成時body identity。",
      "provenance_status": "CANONICAL_READBACK_BODY_HASH_MATCH"
    }
  ],
  "research_sections": [
    {
      "heading": "WHY_THIS_MATTERS_TO_UPDATE",
      "text": "企業AIでは「画像入力できるLLM」と専用Vision API/OCR/動画解析が同じ“vision”として語られやすい。しかし detection / OCR / caption / visual reasoning / moderation / video temporal analysis は異なる能力・version・failure envelopeを持つ。後続Workが用途別に採否・provider mappingできる共通契約が必要。"
    },
    {
      "heading": "SCOPE_ROUTING",
      "text": "Enterprise AI Core / provider mapping / benchmark / implementation / governance / security-privacy / non-LLM component / interoperability。EA-070 Document IntelligenceとはOCR・layout境界で接続するが、文書構造抽出を本項へ吸収しない。"
    },
    {
      "heading": "CLAIM_OR_RESEARCH_QUESTION",
      "text": "1) Vision AIをどの能力単位へ分離すべきか。2) 専用Vision APIとMLLMの出力を同一保証として扱えるか。3) 画像・動画の認識結果、confidence、spatial/temporal locator、model/API revisionをどうprovenance化するか。4) visual reasoning/retrievalのbenchmarkを製品保証へ外挿せず、どのfailure classを運用試験へ移すべきか。"
    },
    {
      "heading": "EVIDENCE TABLE",
      "text": "- official product spec | “What is Image Analysis?” | Microsoft | https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/overview-image-analysis | updated 2025-09-26 / accessed 2026-09-15 | Image Analysis API v4.0/v3.2 | v4.0: Read text, Captions, Dense captions, Tags, Object detection, People, Smart crop; v3.2は別feature set。4.0 background removal/Segment APIは2025-03-31 retired | VERIFIED_SPEC + PRODUCT_VOLATILE | support/context/limit\n- official product overview | “Vision AI: Image and visual AI tools” | Google Cloud | https://cloud.google.com/vision | accessed 2026-09-15 | current web product surface | object detection, image/document/video processing, understanding, search, classification, moderationをvision用途として列挙 | VERIFIED_SPEC + PRODUCT_VOLATILE | context/support\n- official product spec | “Detecting objects and concepts” | Amazon Web Services | https://docs.aws.amazon.com/rekognition/latest/dg/labels.html | accessed 2026-09-15 | Rekognition current docs | labelsはobject/scene/action/concept。gender-like physical appearance predictionはgender identity分類用途ではない旨を明示 | VERIFIED_SPEC + PRODUCT_VOLATILE | support/limit\n- official API/spec | “Using the image and video moderation APIs” | Amazon Web Services | https://docs.aws.amazon.com/rekognition/latest/dg/moderation-api.html | accessed 2026-09-15 | moderation taxonomy L1-L3; ModerationModelVersion returned | confidence threshold、image/video timestamp/segment aggregation、model versionを取得可能。適否判断はapplication policy側 | VERIFIED_SPEC + PRODUCT_VOLATILE | support/limit\n- research benchmark | “Visual Reasoning Benchmark: Evaluating Multimodal LLMs on Classroom-Authentic Visual Problems from Primary Education” | Huti et al. | https://arxiv.org/abs/2602.12196 | 2026-02-12 / accessed 2026-09-15 | 701 questions | static counting/scalingよりfolding/reflection/rotation等でspatial ceilingを報告 | RESEARCH | counter/limit\n- research benchmark | “VisR-Bench: An Empirical Study on Visual Retrieval-Augmented Generation for Multilingual Long Document Understanding” | Chen et al. | https://arxiv.org/abs/2508.07493 | 2025-08-10 / accessed 2026-09-15 | >35K QA, 1.2K docs, 16 languages | MLLM retrievalがtext/multimodal encoderより優位な設定がある一方、structured tablesとlow-resource languagesで課題 | RESEARCH | support/counter/limit"
    },
    {
      "heading": "FINDINGS",
      "text": "A. Vision capabilityを少なくとも (a) OCR/text reading, (b) image classification/tagging, (c) object detection/localization, (d) caption/dense description, (e) visual question answering/reasoning, (f) multimodal embedding/retrieval, (g) moderation/safety classification, (h) video temporal analysis に分離する。単一“vision capable” flagは禁止候補。\nB. 専用Vision APIはfeature taxonomy、confidence、bounding box/timestamp、model version等のstructured outputを持ち得る。一方MLLMの自然言語回答は同じ検出保証ではない。専用detectorのconfidenceとLLMの自己申告confidenceを同一尺度にしない。\nC. version driftは実在する。Azure Image Analysisはv4.0/v3.2でfeature setが異なり、一部featureはretire済み。current product claimはAPI/model revisionを固定する。\nD. moderation labelはpolicy decisionそのものではない。AWSもconfidence thresholdとtaxonomyを返し、application側が適否を決める構造。model output→policy mapping→human/escalationを分離する。\nE. visual benchmarkは能力のfailure classを示すが、商用tenant/modelの保証にはならない。VRBのspatial reasoning weakness、VisR-Benchのtable/low-resource language weaknessはTEST_PROFILE設計根拠に限定する。"
    },
    {
      "heading": "LIMITS / COUNTEREVIDENCE / FAILURE ENVELOPE",
      "text": "- OCR成功≠visual reasoning成功、caption正確≠object localization正確、image accuracy≠video temporal accuracy。\n- bounding box/labelが正しくても関係・数量・左右・前後・遮蔽・rotation・reflection・temporal orderで誤る可能性。\n- MLLM caption/VQAは画像にないobject/text/relationを生成し得るため、high-stakes extractionではsource locatorとdetector/structured evidenceを優先し、unsupported visual claimを監査対象にする。\n- moderation taxonomy/thresholdは文化・policy・use case依存。provider labelを企業policyへ無変換で採用しない。\n- benchmark dataset/model時点とcurrent provider versionを混同しない。"
    },
    {
      "heading": "AVAILABILITY_GATES",
      "text": "provider / product / API version / model ID / region / input type(image, stored video, streaming video) / file format-size-resolution-duration / language / feature flag-preview-GA-retired / quota / pricing tier / data residency。利用可否はfeature存在とは別claim。"
    },
    {
      "heading": "IMPLEMENTATION_STATE",
      "text": "Microsoft Azure Vision Image Analysis v4.0/v3.2、Google Cloud Vision AI product surface、AWS Rekognition Image/Videoはcurrent official implementation evidenceあり。MLLM各providerの最新vision model ID・image/video limits・data controlsは変動が大きいため、本Packetでは横断確定せずprovider mapping再採集対象。"
    },
    {
      "heading": "BENCHMARK_OR_EVAL",
      "text": "後続TEST_PROFILE（未実施）: 同一fixtureを専用Vision APIとMLLMへ入力。classes: clean/noisy text image, multi-object, small/occluded object, counting, spatial relation, rotation/reflection, chart/table screenshot, UI screenshot, multilingual image, unsafe-content edge case, short video with temporal event。metrics: OCR CER/WER, classification precision/recall/F1, localization IoU/mAP相当, count exact match, relation/spatial accuracy, caption factual precision, unsupported-object/relation rate, VQA accuracy, retrieval Recall@K/nDCG, moderation precision/recall at threshold, temporal event/timestamp error, abstention, latency, cost/image-or-minute。benchmark scoreをtenant保証にしない。"
    },
    {
      "heading": "SECURITY/PRIVACY",
      "text": "画像・動画は顔、画面、文書、位置情報等を含み得る。入力原本、派生caption/embedding/label、保存期間、downstream logを別artifactとして扱う。人物属性推定は用途制限・法務/倫理review対象。AWS docsのgender-like appearance labelもgender identity推定用途ではない。\nAUDIT: source hash/URI, frame/page/time locator, crop/ROI, model/API revision, feature, threshold, raw structured output, transformed prompt/context, final claim, human overrideを追跡。"
    },
    {
      "heading": "COST/OPERABILITY",
      "text": "per-image/page/minute/API-call/token等meterが異なるため、MLLM token costとVision API unit costを分離。動画は非同期job・pagination・segment aggregation等も運用設計対象。"
    },
    {
      "heading": "SAFE_WORDING_FOR_FUTURE_BOOK",
      "text": "「Vision AIは単一能力ではない。OCR、検出、localization、caption、visual reasoning、retrieval、moderation、video temporal analysisを能力単位で評価し、model/API revisionとsource locatorを保持する。専用Vision APIのstructured detector outputとMLLMの自然言語理解は補完関係にあり、片方の成功を他方の保証へ外挿しない。」"
    },
    {
      "heading": "UPDATE_WORK_HANDOFF",
      "text": "Core候補 `VISION EVIDENCE CONTRACT` を作成する: source_id/hash; media_type; page/frame/timestamp/ROI; task_type; provider/product/model/API revision; preprocessing; label/text/object/relation; geometry; confidence; moderation taxonomy+threshold; raw output locator; derived caption/embedding; principal/ACL; retention; final claim; human override。EA-070 Document Extraction ContractとOCR/page geometryで接続し、document semanticsはEA-070側を正とする。provider mapping時はcurrent model IDs、region/data residency、image/video limits、pricing、retentionを再確認。"
    },
    {
      "heading": "OPEN_UNKNOWN",
      "text": "/ STOP_CONDITION"
    },
    {
      "heading": "UNKNOWN",
      "text": "2026-09-15時点の主要MLLM各社についてvision/video model ID・最大media constraints・region/data-controlを同一比較軸で一次資料固定していない。各provider mapping時にPRODUCT_VOLATILEとして再採集する。STOP_CONDITION: capability taxonomy、3 providerの実装例、version/retirement evidence、反証benchmark、future contract、safe test profileが揃ったためEA-072 bounded packetは終了。再調査は新しいmajor API/model revision、retirement、または後続Workが特定provider claimを必要とした場合のみ。\n\n---"
    }
  ],
  "human_url": "/tools/ea-072-vision-video-ai",
  "markdown_url": "/agent/article.md?path=%2Ftools%2Fea-072-vision-video-ai",
  "json_url": "/agent/article.json?path=%2Ftools%2Fea-072-vision-video-ai"
}