Configuring the label editor

Once you've created a dataset, you can further configure the labeling interface by editing the configuration file under Settings -> Labeling.

You can also copy-paste this configuration file from one dataset to another.

{
"format_version": "0.1",
"categories": [
{ // At the minimum, a category should have a name and category_id
"name": "person",
"category_id": 1
},
{
"name": "car",
"category_id": 2,
"has_instances": true, // Whether category contains instances (person, car) or not (sky, road)
"color": [33, 138, 33], // RGB color of the category
"foo": "bar" // You can add custom key-value pairs that will be ignored
},
{
"name": "traffic-light",
"category_id": 3,
"attributes": [ // Optional object-level attributes
{
"name": "color",
"input_type": "radio",
"values": [
"green",
"yellow",
"red",
]
},
{
"name": "visibility",
"input_type": "radio",
"values": [
"good",
"bad"
]
}
]
}
],
"image_attributes": [ // Optional image-level attributes
{
"name": "scene_type",
"input_type": "radio",
"values": [
"highway",
"urban",
"countryside"
]
},
{
"name": "weather",
"input_type": "radio",
"values": [
"sunny",
"cloudy",
"rainy"
]
}
]
}