{
  "docId": "019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "docSlug": "bf350dd574c19997",
  "documentTitle": "2021 Air Street Capital The State of AI Report 2021",
  "authorId": "AirStreetCapital",
  "authorName": "Air Street Capital",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.777,
  "pageNumber": 13,
  "pageCount": 188,
  "prevPage": 12,
  "nextPage": 14,
  "slideType": "case_study",
  "function": "illustrate_case",
  "density": "dense",
  "nDataPoints": 1,
  "notes": "The slide showcases visual evidence of unsupervised object segmentation via attention maps.",
  "elementsJson": [
    "headline_text",
    "bullet_list",
    "photo"
  ],
  "metadataConfidence": 1,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-fd5012384ae3/13",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3#slide-13",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "By inspecting the self-attention module of the last block of SSViTs, the authors show that SSViTs learn \"class-specific features leading to unsupervised object segmentations\".",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d24-9ab579a54edf",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.28,
        "w": 0.58,
        "x": 0.37,
        "y": 0.58
      },
      "kind": "image",
      "text": "Grid of images showing original photos paired with their corresponding attention maps highlighting objects.",
      "attrs": null,
      "subkind": "infographic",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "42cdabb9-c5ac-44ec-92c2-5e5a734c4b35",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.55,
        "w": 0.35,
        "x": 0.05,
        "y": 0.3
      },
      "kind": "list",
      "text": "By inspecting the self-attention module of the last block of SSViTs, the authors show that SSViTs learn \"class-specific features leading to unsupervised object segmentations\". The features learned by SSViTs are very powerful: They achieve 78.3% top-1 accuracy on ImageNet when using these features and a simple k-NN algorithm without fine-tuning or data augmentation. They show that these properties don't emerge for supervised ViTs and convnets. They also compare to other self-supervised methods and a supervised ViT trained on ImageNet, and show that a self-supervised ViT outperforms them on a video segmentation task.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "3d6b5ddc-97bf-499a-868e-84d702a5065c",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "top-1 accuracy: 78.3%",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d24-9c0b7fda7cb2",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.9,
        "x": 0.05,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "Researchers compare a self-supervised ViT (SSViT) to fully supervised ViTs and convnets, and find that SSViTs learn more powerful representations.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "60632d44-48f4-417d-9c2b-379976a20a53",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.8,
        "x": 0.05,
        "y": 0.13
      },
      "kind": "title",
      "text": "What do self-supervised Vision Transformers see in an image that other models don't?",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f8db3355-474f-4b8d-9574-6fcc60baeb85",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Information Gap Theory",
      "slug": "information-gap-theory",
      "agent": "Storyteller",
      "layer": "loop",
      "matchId": "019dd95a-0fd5-7148-8ed0-cb1ca1be3725",
      "evidence": "Title is a curiosity-gap question 'What do SSViTs see…'",
      "confidence": 65
    },
    {
      "name": "Action Titles",
      "slug": "action-titles",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "019dd95a-0fd5-7148-8ed0-c64b2043952c",
      "evidence": "Title poses the question + answer pattern.",
      "confidence": 70
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 152,
      "from": 5,
      "beatId": "019dd95a-0682-776c-8e35-0affbadec38e",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Definitions p.5-6, Exec Summary p.7, then Sections 1-3 catalog evidence.",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 152,
      "from": 10,
      "beatId": "019dd95a-0682-776c-8e35-1b3ee3a1c012",
      "arcName": "The Mountain",
      "arcSlug": "mountain",
      "beatName": "Rising Action",
      "beatSlug": "mountain-rising-action",
      "evidence": "Three sections accumulating evidence of accelerating progress.",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 18,
      "from": 11,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-45052222b756",
      "evidence": "Multiple evidence slides (vision, audio, 3D, MLPs comeback) all converging on transformer dominance.",
      "position": 2,
      "objective": "Show transformers spreading across AI domains",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 82,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}