Skip to content

Latest commit

 

History

History
145 lines (116 loc) · 5.36 KB

README.md

File metadata and controls

145 lines (116 loc) · 5.36 KB
Raw

Export As Masks

Overview • How To Run • How To Use

GitHub release (latest SemVer) views runs

Overview

Export prepares downloadable .tar archive, that contains:

  • original images
  • annotations in Supervisely JSON format
  • human masks - RGB masks where every pixel has the color of the corresponding class (semantic segmentation)
  • machine masks. Notice: if you open machine mask image in standard image viewer, it will look like completely black image, but it is not. Class colors for machine mask are generated automatically as indices of classes. (0, 0, 0) - is always a background (unlabeled area), (1, 1, 1) - for class #1, (2, 2, 2) - for class #2, etc ... Mapping between machine colors and classes in machine mask is saved in obj_class_to_machine_color.json file.
  • instance masks - BW masks for every object on the image (instance segmentation)

🏋️ Starting from version v2.1.11 application supports split archives. If the archive file size is too big, it will be split into several parts. Learn more below in the "How to extract split archives" section.

For example:

{
  "kiwi": [1, 1, 1],
  "lemon": [2, 2, 2]
}

Output example:

<id_project_name>.tar
├── cat
│   ├── ann
│   │   ├── cats_1.jpg.json
│   │   ├── ...
│   │   └── cats_9.jpg.json
│   ├── img
│   │   ├── cats_1.jpg
│   │   ├── ...
│   │   └── cats_9.jpg
│   ├── masks_human
│   │   ├── cats_1.png
│   │   ├── ...
│   │   └── cats_9.png
│   ├── masks_instances
│   │   ├── cats_1
│   │   │   ├── cats_1.png
│   │   │   ├── ...
│   │   │   └── cats_9.png
│   │   ├── ...
│   │   └── cats_9
│   │       ├── cats_1.png
│   │       ├── ...
│   │       └── cats_9.png
│   └── masks_machine
│       ├── cats_1.png
│       ├── ...
│       └── cats_9.png
└── dog
    ├── ann
    │   ├── dogs_1.jpg.json
    │   ├── ...
    │   └── dogs_9.jpg.json
    ├── img
    │   ├── dogs_1.jpg
    │   ├── ...
    │   └── dogs_9.jpg
    ├── masks_human
    │   ├── dogs_1.png
    │   ├── ...
    │   └── dogs_9.png
    ├── masks_instances
    │   ├── dogs_1
    │   │   ├── dogs_1.png
    │   │   ├── ...
    │   │   └── dogs_9.png
    │   ├── ...
    │   └── dogs_9
    │       ├── dogs_1.png
    │       ├── ...
    │       └── dogs_9.png
    └── masks_machine
        ├── dogs_1.png
        ├── ...
        └── dogs_9.png

How To Run

Step 1: Add app to your team from Ecosystem if it is not there

Step 2: Open context menu of images project -> Download via App -> Export as masks

Step 3: Define export settings in modal window

Step 4: Result archive will be available for download:

  • single archive: in the Tasks list (image below) or from Team Files
    • Team Files->tmp->supervisely->export->export-as-masks->task_id-><projectId>_<projectName>.tar
  • split archive: all parts will be stored in the Team Files directory
    • Team Files->tmp->supervisely->export->export-as-masks-><task_id>

How to extract split archives

In the case of a split archive:

  1. download all parts from Team Files directory (Team Files->tmp->supervisely->export->export-as-masks-><task_id>)
  2. After downloading all archive parts, you can extract them:

  • for Windows: use the following freeware to unpack Multi-Tar files: 7-zip and click on the first file (with extension .tar.001)

  • for MacOS: replace <path_to_folder_with_archive_parts>, <projectId> and <projectName> with your values and run the following commands in the terminal:

cd <path_to_folder_with_archive_parts>
cat <projectId>_<projectName>.tar* | tar --options read_concatenated_archives -xvf -
  • for Linux (Ubuntu): replace <path_to_folder_with_archive_parts>, <projectId> and <projectName> with your values and run the following commands in the terminal:
cd <path_to_folder_with_archive_parts>
cat '<projectId>_<projectName>.tar'* > result_archive.tar | tar -xvf result_archive.tar