cm-tup

Scripts to create CVAT or Segments.AI tasks from a directory of tiles.

Setup for CVAT and segments.ai

1. Clone the CVAT repository if you don't have it yet.

    git clone https://github.com/kappazeta/cvat.git
   

2. Enter the directory of the cm-tup repository.

3. Create a conda environment for the CVAT CLI utility.

    conda env create -f environment.yml
   

Configuration

The script expects an S2 product directory which has been subtiled and converted, using cm-vsm (https://github.com/kappazeta/cm-vsm.git).

In the cm-vsm output directory, cm-tup expects to find a configuration file cvat-tup.cfg and a label classes definition file classes.json. In the case that these files don't exist yet, cm-tup creates them from templates.

An example cvat-tup.cfg file:

cvat_host_addr=localhost
cvat_host_port=8080
cvat_credentials=YOUR_CVAT_USERNAME:YOUR_CVAT_PASSWORD
cvat_cli_path=/YOUR_PATH_TO_CVAT_WORKING_COPY/utils/cli/cli.py
git_repository=https://github.com/kappazeta/cvat_cloudmask_annotations.git

Here, YOUR_PATH_TO_CVAT_WORKING_COPY, YOUR_CVAT_USERNAME and YOUR_CVAT_PASSWORD ought to be replaced with values which correspond to your setup.

An example classes.json file:

[
    {   
        "name": "CLEAR",
        "attributes": []
    },  
    {   
        "name": "CLOUD_SHADOW",
        "attributes": []
    },  
    {   
        "name": "CLOUD",
        "attributes": []
    },  
    {   
        "name": "SEMI_TRANSPARENT_CLOUD",
        "attributes": []
    }   
]

Running

1. Activate the conda environment

    source activate cvat-cli
   

2. Run the script with the output directory of cm-vsm as the only argument. Note that the path should end with the suffix .CVAT.

    ./bin/cvat-tup.sh YOUR_PATH
   

Segments.ai configuration and running

1. First, you need to have an user account on https://segments.ai
2. Under bin directory, segments-ai-uploader.py is used to upload tiles in your dataset. It should be configured in the following way:

• #unique_client_id should be replaced with a newly generated API key from your segments.ai user profile;

• #dataset_name should be replaced with a full dataset name (e.g. your_user_name/playground);

• #path_to_directory_with_subtiles should be replaced with a full path to a directory with subtiles. By default the script expects the folder to be a collection of products with tiles inside (e.g. target_directory/product_name/tile_x_y.png). To change the nesting level and search for tiles right inside the specified directory /**/*.png can be changed to /*.png, where * is a wildcard character.

3. When the configuration is done, run the script by typing python segments-ai-uploader.py under cvat-cli conda environment.