Thumbnail extractor

Example pipeline for tator online. Can be used to extract png images of annotated boxes from video and images. Uploads result to a new image.

Required pipeline arguments:

imageTypeId : the type id to use for the thumbnail image

Sample config.json

config.json can be used with the tator_testHarness.py script

tator_testHarness.py config.json tator/setup.py

  "TATOR_AUTH_TOKEN": "<TOKEN>",
  "TATOR_PROJECT_ID": "1",
  "TATOR_API_SERVICE": "https://debug.tatorapp.com/rest",
  "TATOR_MEDIA_IDS": "1",
  "TATOR_PIPELINE_ARGS" : "{\"imageTypeId\": 10}"
}

Example from Tator Online

The arguments attribute in an Algorithm model could look like the following:

{"mode": "localization_keyframe",
"type_id": 48,
"imageTypeId": 50,
"dest_section": "Training Box Extracts"}

Where:

mode: One of

• localization_keyframe: Extracts whole frame duplicates localizations in new image

• localization_thumbnail: Extracts content of localization box and applies attributes as media attributes to the resultant image.

• state: Same as localization_keyframe but with State types.

type_id: The underlying type of the object to base extractions off of.

imageTypeId: The destination image type.

dest_section: The name of the section to place the extracted imagery.

Testing

make test unit test make docker_test use tator test harness and invoke against server