Collection Campaign Scripting
You can now use the Python client to grab the formatted results from either your (a) unlabeled collection campaigns in the UI or (b) more traditional Python client collection campaign.
Full python client docs are available here, and a brief example follows.
import aquariumlearning as al
import json
# Your issue must have been created under the Rare Scenario issue type
ISSUE_UUID = "..."
# Either the unlabeled dataset that you want to search through
# OR "" if you are querying the results of a Python-client-triggered
# collection campaign
collection_client = al.CollectionClient()
results = collection_client.get_unlabeled_results_for_issue(
# You can add the option retrigger_search=True to run the search
# for the first time (or rerun)
print(f"{len(results)} results found")
with open(OUTPUT_JSON, "w") as f:
json.dump(results, f)
A particular unlabeled result might look something like the following (note: this is for the bounding box objection detection task):
"classifierScore": 0.8154874556308199,
"collectionFrameId": 536914,
"deviceId": "default_device",
"images": [
"coordinateFrame": "img0",
"dateCaptured": "2022-01-04T14:56:34.384042+00:00",
"imageBoxes": [
"height": 39.599999999999994,
"label": "car",
"left": 741.18,
"top": 168.83,
"user__width": 51.07000000000005,
"user__width_bucket": "medium",
"uuid": "000008_8_1_gt",
"width": 51.07000000000005,
"windowId": 1641326258849
"height": 16.580000000000013,
"label": "dontcare",
"left": 826.87,
"top": 162.28,
"user__width": 18.970000000000027,
"user__width_bucket": "small",
"uuid": "000008_8_2_gt",
"width": 18.970000000000027,
"windowId": 1641326258849
"image_url": "",
"preview_url": "",
"sensorId": "img0"
"issueVersion": 1,
"labeled": false,
"sampledElementId": "000008_8_2_gt",
"similarityScore": 0.8673680939432816,
"sortStatus": "UNSORTED",
"sourceMetadata": {
"source": "unlabeled_dataset",
"source_dataset": "test_pca_diff.unlabeled_real_str"
"taskId": "000008_8"
Copy link