Cityscapes

Overview

This converter allows to import images with .json annotations in Cityscapes format.

⚠️ Note: images must have suffix _leftImg8bit and annotations suffix _gtFine_polygons and .json extension. Check the example of the file structure below.

Result of the import

Enterprise users have access to "Import as links" option, which supports import of this format with annotations. This option might be beneficial in many cases, as it allows data import to Supervisely platform without re-uploading, maintaining a single source and speeding up import process.

To step up import speed even further you can compress all annotation files (.json's) into an archive and import it together with the images. (Note: This method is format-dependent and may not apply to all formats.)

Format description

Supported image formats: .jpg, .jpeg, .mpo, .bmp, .png, .webp, .tiff, .tif, .jfif, .avif, .heic, and .heif With annotations: yes Supported annotation file extension: .json. Grouped by: any structure (uploaded to a single dataset)\

Input files structure

Example data: download ⬇️\

Recommended directory structure:

📦project name
 ┣ 📂gtFine
 ┃ ┣ 📂test
 ┃ ┃ ┗ 📂ds1
 ┃ ┃ ┃ ┗ 📜IMG_8144_gtFine_polygons.json
 ┃ ┣ 📂train
 ┃ ┃ ┗ 📂ds1
 ┃ ┃ ┃ ┣ 📜IMG_1836_gtFine_polygons.json
 ┃ ┃ ┃ ┣ 📜IMG_2084_gtFine_polygons.json
 ┃ ┃ ┃ ┣ 📜IMG_3861_gtFine_polygons.json
 ┃ ┃ ┃ ┗ 📜IMG_4451_gtFine_polygons.json
 ┃ ┗ 📂val
 ┃ ┃ ┗ 📂ds1
 ┃ ┃ ┃ ┗ 📜IMG_0748_gtFine_polygons.json
 ┣ 📂leftImg8bit
 ┃ ┣ 📂test
 ┃ ┃ ┗ 📂ds1
 ┃ ┃ ┃ ┗ 🖼️IMG_8144_leftImg8bit.png
 ┃ ┣ 📂train
 ┃ ┃ ┗ 📂ds1
 ┃ ┃ ┃ ┣ 🖼️IMG_1836_leftImg8bit.png
 ┃ ┃ ┃ ┣ 🖼️IMG_2084_leftImg8bit.png
 ┃ ┃ ┃ ┣ 🖼️IMG_3861_leftImg8bit.png
 ┃ ┃ ┃ ┗ 🖼️IMG_4451_leftImg8bit.png
 ┃ ┗ 📂val
 ┃ ┃ ┗ 📂ds1
 ┃ ┃ ┃ ┗ 🖼️IMG_0748_leftImg8bit.png
 ┗ 📜class_to_id.json

Format Config File

In order to import custom annotations for the images, you need to provide a class_to_id.json file. This file should contain a list with dictionaries. Each dictionary should contain information about the class with the following fields:

• name - the name of the class. It should be unique.

• id - the ID of the class. From 1 to N-1, where N is the number of classes.

• color - the color of the class in RGB format. If not specified, the color will be generated randomly

📜class_to_id.json

[
  {
    "name": "kiwi",
    "id": 1,
    "color": [255, 0, 0]
  },
  {
    "name": "lemon",
    "id": 2,
    "color": [81, 198, 170]
  }
]

Single-Image Annotation JSON

Annotation file should contain the following fields:

• imgHeight - the height of the image

• imgWidth - the width of the image

• objects - a list of dictionaries, each containing information about the object

  • label - the name of the class

  • polygon - a list of points that form the polygon of the object

Example of the annotation file from provided sample data:

📜IMG_1836_gtFine_polygons.json

{
    "imgHeight": 800,
    "imgWidth": 1067,
    "objects": [
        {
            "label": "lemon",
            "polygon": [
                [772, 421],
                [771, 422],
                ...
                [785, 422],
                [784, 421]
            ]
        },
        {
            "label": "kiwi",
            "polygon": [
                [637, 122],
                [636, 123],
                ...
                [645, 123],
                [644, 122]
            ]
        },
        {
            "label": "kiwi",
            "polygon": [
                [543, 539],
                [542, 540],
                ...
                [548, 540],
                [547, 539]
            ]
        }
    ]
}

Useful links

• Cityscapes format

• [Supervisely Ecosystem] Import Cityscapes