Theos JSON

Our official bounding box label format.

Overview

In the Theos JSON format, each dataset image has its own labels file. For example, if we have an image named image.jpg we should also have an image.json file where the labels in this image will be stored. When you import a folder into Theos, all images and labels within all subdirectories must have a unique name, otherwise they will be incorrectly recognized as the same image.

Folder structure

• classes.json

• train

  • images

    • image1.jpg

    • image2.jpg

    • image(N).jpg

  • labels

    • image3.json

    • image4.json

    • image(N+1).json

• valid

  • images

    • image5.jpg

    • image6.jpg

    • image(N+2).jpg

  • labels

    • image5.json

    • image6.json

    • image(N+2).json

• test

  • images

    • image7.jpg

    • image8.jpg

    • image(N+3).jpg

  • labels

    • image7.json

    • image8.json

    • image(N+3).json

The classes file

This is where all the classes of the dataset are defined. Each class is composed of its id, name, color, whether if it is_superclass or not (is_superclass should always be false) and whether if it's a class with_text or not (for OCR labeling). The first id must always be 0, and consequent ids must be in order: 0, 1, 2, 3, 4, 5, etc.

Following is an example of a classes.json file.

classes.json

[
  {
    "id": 0,
    "name": "eye",
    "color": "#50e3c2",
    "is_superclass": false,
    "with_text": false
  },
  {
     "id": 1,
     "name": "nose",
     "color": "#e155f8",
     "is_superclass": false,
     "with_text": false
  },
  {
     "id": 2,
     "name": "mouth",
     "color": "#f66173",
     "is_superclass": false,
     "with_text": false
  },
  {
     "id": 3,
     "name": "face",
     "color": "#8142f5",
     "is_superclass": false,
     "with_text": false
  }
]

The label file

This is a file representing all the labels present within a particular image.

einstein.jpg

einstein.jpg

einstein.json

Each bounding box is composed of its class_id, (x, y) position of its top left point, and its width and height.

einstein.json

[
  {
    "class_id": 0,
    "x": 687,
    "y": 579,
    "width": 65,
    "height": 38
  },
  {
    "class_id": 0,
    "x": 498,
    "y": 575,
    "width": 81,
    "height": 42
  },
  {
    "class_id": 1,
    "x": 586,
    "y": 579,
    "width": 122,
    "height": 198
  },
  {
    "class_id": 2,
    "x": 563,
    "y": 835,
    "width": 153,
    "height": 48
  },
  {
    "class_id": 3,
    "x": 347,
    "y": 324,
    "width": 453,
    "height": 660
  }
]

If you are also labeling for OCR (optical character recognition) you must add a text field.

vehicle.json

[
  {
    "class_id": 0,
    "x": 687,
    "y": 579,
    "width": 65,
    "height": 38,
    "text": "6XSU832"
  }
]