Export Supervisely to Cityscapes
Converts Supervisely project to Cityscapes format and prepares downloadable tar archive.
Supervisely project have to contain only classes with shapes Polygon or Bitmap. It means that all labeled objects have to be polygons or bitmaps. If your project has classes with other shapes and you would like to convert the shapes of these classes and all corresponding objects (e.g. rectangles to polygons), we recommend to use Convert Class Shape app.
1) Labeled foreground objects must never have holes(Cityscapes format doesn’t support them), i.e. if there is some background visible through a foreground object, it is considered to a be part of the foreground. If objects in Supervisely format have holes, after export to Cityscapes format holes will be filled with class color, so you may lose some of the input data.
2) If Bitmap in Supervisely format contain unconnected components, this figure will be splitted into separate objects after export to Cityscapes format.
3) Unlike the standard Cityscapes format, which can contain well-defined classes with well-defined colors and ids, project derived from Supervisely format can contain classes with any names and colors.
In this regard:
JSON format will be added to the root of the resulting project with name class_to_id.json. It contains information about Supervisely project classes names, colors and their corresponding ids. <image_name>_instanceIds.png will not be created after conversion(information about classes colors and ids is contained in class_to_id.json file)Cityscapes dataset structure
The folder structure of the Cityscapes dataset is as follows:
{root}/{type}{video}/{split}/{city}/{city}_{seq:0>6}_{frame:0>6}_{type}{ext}
The meaning of the individual elements is:
root the root folder of the Cityscapes dataset.
type the type/modality of data, e.g. gtFine for fine ground truth, or leftImg8bit for left 8-bit images.
split the split, i.e. train/val/test. Note: not all kinds of data exist for all splits. Thus, do not be surprised to occasionally find empty folders.
city location(city) where part of the dataset was recorded. In Supervisely project city is used as a dataset name.
You can download example of Cityscapes datasets here.
Current version of application supports only:gtFine - the fine annotations. This type of annotations is used for validation, testing, and optionally for training. Annotations are encoded using json files containing the individual polygons.leftImg8bit - the images. These are the standard annotated images.
In addition, Cityscapes format implies the presence of train/val datasets, and also train/val/test. Thus, to split images on training and validation datasets you should assign corresponding tags (train, val, test) to images. If project contains one of train, val, test tags, resulting archive also will contain it. If image doesn’t have such tags, tags train, val and test will be assigned automatically in proportions train(60%), val(20%), test(20%).
Step 1: Add app to your team from Ecosystem if it is not there.
Step 2: Open context menu of project -> Download as -> Export to Cityscapes format
After running the application, you will be redirected to the Tasks page. Once application processing has finished, your link for downloading will become available. Click on the file name to download it.
Note You can also find your converted project in: Team Files->cityscapes_format->app_id->projectId_projectName_formatName.tar.gz