Asset metadata format

Assets need to be registered.

To register an asset group, you need to prepare a metadata json file with information about the asset group.

Next, you need to specify the list of asset ids. This can be specified as a csv file or as a solr index url.

Current asset metadata format (v0.2)

Each asset will have a set of various data associated with it (different views of the given asset). For instance, a model can have associated with it the following types of data:

  mesh: 3D surface mesh representing the geometry and material of the model
  segment: A segmentation of the 3D mesh geometry into parts or regions
  voxel: Voxelization of the 3D mesh geometry into little voxels
  image: screenshots/renderings of the model
  video: recorded videos of the model

To accommodate the different types of data, the metadata format support specifying the data associated with each type in one of the following ways:

1. A flat array with the dataType specified for each element

  "data": [{
    "name": "ply",
    "format": "ply",
    "dataType": "mesh",
    "path": ...
  },
  {
    "name": "surfaces",
    "dataType": "segment",
    "path": ... 
  }]

2. A object keyed by the type This allows for data of the same type to be grouped together and for defaultOptions to be specified for each type. If there are no defaultOptions for the type, the data field can be directly specified as an array.

  "dataTypes": {
    // Specifying default mesh options (used when loading the mesh)
    "mesh": {
       "defaultOptions": {
          ...
       },
       "data": [{
          "name": "ply",
          "format": "ply",
          "path": ...
       }]
    },
    // Specifying "data" array directly
    "segment": [{
       "name": "surfaces",
       "path": ...
    }]
  }

Asset metadata format v0.1

The older asset metadata format (used for the ScanNet annotation tools) assumed that the primary assets data was a mesh, and used names of the field to identify the type of the data. This limited the different versions for a particular data type (for instance, the set of voxels that were associated with an asset).

Example metadata v0.1 json file

{
  "source": "xxx",
  "assetType": "model",                       
  "rootPath": "${baseUrl}/data/xxx/",
  "defaultImageCount": 14,
  "screenShotPath": "${rootPath}/object_screenshots/${id}/${id}-${index}.png",
  "hasThumbnails": true,
  "originalScreenShotPath": "${rootPath}/objects/${baseVariantId}/${id}.png",
  "texturePath": "${rootPath}/texture/",
  "prefetchModelInfo": { "source": "xxx", "query": "*:*", "limit": 4000 },
  "voxels-solid": "${rootPath}/object_vox/object_vox_data/${baseVariantId}/${id}.binvox",
  "assetFields": ["parts"],
  "surfaces": {
    "file": "${rootPath}/object_segmentations/${id}/${id}-segments.json",
    "format": "indexedSegmentation"
  },
  "parts": {
    "file": "${rootPath}/object_part_annotations/${id}/${id}.json",
    "format": "indexedSegmentation"
  },
  "part-annotations": {
    "name": "parts",
    "format": "indexedSegmentation",
    "files": {
      "annIds": "${baseUrl}/part-annotations/list?itemId=${fullId}&$columns=id,workerId,data&format=json",
      "annotation": "${baseUrl}/query?qt=parts&annId=${annId}&modelId=${fullId}",
      "segmentation": "${rootPath}/object_segmentations/${id}/${id}-segments.json"
    }
  },
  "formats": [
    { 
      "name": "kmz",
      "format": "kmz",
      "path": "${rootPath}/objects/${baseVariantId}/${id}.kmz",
      "defaultUp": [ 0, 1, 0 ], "defaultFront": [ 0, 0, 1], "defaultUnit": 1,
      "voxelsAligned": true,
      "voxelsToModelTransform": [1, 0, 0, 0, 0, 1, 0, 1, 0]
    },
  ]
}