Skip to content

Latest commit

 

History

History
81 lines (62 loc) · 1.28 KB

SAMPLE_SPEC.md

File metadata and controls

81 lines (62 loc) · 1.28 KB
Raw

Sample Spec

Every JSON dataset has a samples array containing each unit of work to be labeled.

Not every interface can accept every type of data. Check the interface specification to determine how exactly to format the sample.

Images

{
  "imageUrl": "https://example.com/image1.png"
}

Audio

{
  "audioUrl": "https://example.com/audio.mp3"
}

Video

{
  "videoUrl": "https://example.com/video.mp4"
}

Text / Markdown / Documents

{
  "document": "some string"
}
{
  "markdown": "# title\nSome **markdown**."
}
{
  "textUrl": "https://example.com/example.txt"
}

Time Data

The unit of "time" can be either a duration (time since start) or a unix epoch time. The representation of time is configured in the interface.

{
  "timeData": [
    { "time": 0, "value": 0 }, // Keys other than "value" are fine as well
    { "time" : 500, "value": 0.75 },
    { "time" : 1000, "value": 1 }
  ]
}

If a time interface is selected, each row of the CSV URL will become an object in timeData. The header will be used to determine the keys of each JSON object.

{
  "csvUrl": "https://example.com/example.csv"
}

PDFs

{
  "pdfUrl": "https://example.com/document.pdf"
}