{"help": "https://data.gov.au/data/en/api/3/action/help_show?name=package_show", "success": true, "result": {"archived": false, "author": "Friends of Bool and Hacks Lagoons", "author_email": null, "contact_point": "bradley.clarke-wood@birdlife.org.au", "creator_user_id": "c2fbbe4a-4ba0-4945-808b-67454605a4cf", "duplicate_score": 2, "geospatial_topic": ["environment"], "id": "e99ccbc7-ff9f-4ccc-b20e-ebaee2239189", "isopen": false, "language": "eng", "license_id": "cc-by", "license_title": "Creative Commons Attribution 3.0 Australia", "license_url": "http://creativecommons.org/licenses/by/3.0/au/", "maintainer": "bradley-clarke-wood-3698", "maintainer_email": null, "metadata_created": "2025-12-20T04:33:22.543176", "metadata_modified": "2025-12-20T04:33:28.362705", "name": "bool-and-hacks-lagoons-landcover-model-output-2022", "notes": "The Friends of Bool and Hacks Lagoon group and BirdLife Australia provided Lynker Analytics with aerial photography for the Bools and Hacks lagoon. The imagery consisted of 66 ECW files which covered the Bool lagoon at a resolution of 0.106m. In addition to the imagery, twenty-seven ground truth points were also provided to assist in the correct annotation of the eight target classes.\nLynker then manually annotated these images into a polyline annotation dataset. The classes followed by their class id are:\n\u2022 Tussock 1\n\u2022 Tree 2\n\u2022 Sedge 3\n\u2022 Reed 4\n\u2022 Grasses 5\n\u2022 Open Water 6\n\u2022 Ground 7\n\u2022 Aquatic Floating 8\nLynker used a machine learning training process called supervised learning, whereby a machine learning model is trained using example image and annotation pairs to learn the same decision outcomes on new or previously unseen images.\nMachine Learning is notoriously data-hungry and model accuracy is sensitive to the quality and quantity of input data. An annotation process that used polylines to quickly develop a large dataset of positively annotated pixels was used to develop the dataset of target classes to train the supervised model.\n\nThe model\u2019s performance on holdout data was shown to have a classification accuracy of 0.965 and mean F1 score also of 0.965. Sedge was the lowest performing class often instead being predicted to be grasses or ground. The aquatic floating class was the highest performing class in the holdout set, every pixel of this class in the holdout set was correctly predicted and no other classes were incorrectly predicted to belong to the aquatic floating class.\n\nSee raw imagery here: https://data.sa.gov.au/data/dataset/a908a10b-b3c0-40e2-a2f9-0ed3849579c7", "num_resources": 1, "num_tags": 8, "organization": {"id": "e1154665-d6b9-412f-8ba2-22443e7d9b8c", "name": "friends-of-bool-and-hacks-lagoons-south-australia-government", "title": "Friends of Bool and Hacks Lagoons", "type": "organization", "description": "The Friends of Bool and Hacks Lagoons is the independent body \r\nestablished to conserve and monitor Bool and Hacks Lagoons.", "image_url": "", "created": "2025-12-20T14:51:48.927295", "is_organization": true, "approval_status": "approved", "state": "active"}, "original_harvest_source": {"site_url": "https://data.sa.gov.au/data/", "href": "https://data.sa.gov.au/data/dataset/bool-and-hacks-lagoons-landcover-model-output-2022", "title": "South Australia Government"}, "original_name": "bool-and-hacks-lagoons-landcover-model-output-2022", "owner_org": "e1154665-d6b9-412f-8ba2-22443e7d9b8c", "private": false, "promotion_level": "0", "remote_last_updated": "2023-05-17 02:01:35.766724", "spatial": "{\"type\": \"Point\",\"coordinates\": [135.866666,-29.383333]}", "spatial_coverage": "SA0062407: South Australia", "state": "active", "temporal_coverage_from": "2022-12-02 00:00:00", "temporal_coverage_to": "2022-12-02 00:00:00", "title": "Bool and Hacks Lagoons Landcover Model Output 2022", "type": "dataset", "unpublished": false, "url": null, "version": null, "extras": [{"key": "harvest_object_id", "value": "3c7a4728-6cce-4401-9a1a-76da1630dacb"}, {"key": "harvest_source_id", "value": "bb56156a-72e4-461b-bea4-aec9145d8b5c"}, {"key": "harvest_source_title", "value": "South Australia Government"}], "resources": [{"cache_last_updated": null, "cache_url": null, "created": "2023-05-17T01:27:53.235783", "datastore_active": false, "datastore_contains_all_records_of_source_file": false, "description": "The Friends of Bool and Hacks Lagoon group and BirdLife Australia provided Lynker Analytics with aerial photography for the Bools and Hacks lagoon. The imagery consisted of 66 ECW files which covered the Bool lagoon at a resolution of 0.106m. In addition to the imagery, twenty-seven ground truth points were also provided to assist in the correct annotation of the eight target classes.\r\nLynker then manually annotated these images into a polyline annotation dataset. The classes followed by their class id are:\r\n\u2022 Tussock 1\r\n\u2022 Tree 2\r\n\u2022 Sedge 3\r\n\u2022 Reed 4\r\n\u2022 Grasses 5\r\n\u2022 Open Water 6\r\n\u2022 Ground 7\r\n\u2022 Aquatic Floating 8\r\nLynker used a machine learning training process called supervised learning, whereby a machine learning model is trained using example image and annotation pairs to learn the same decision outcomes on new or previously unseen images.\r\nMachine Learning is notoriously data-hungry and model accuracy is sensitive to the quality and quantity of input data. An annotation process that used polylines to quickly develop a large dataset of positively annotated pixels was used to develop the dataset of target classes to train the supervised model.\r\nThe model\u2019s performance on holdout data was shown to have a classification accuracy of 0.965 and mean F1 score also of 0.965. Sedge was the lowest performing class often instead being predicted to be grasses or ground. The aquatic floating class was the highest performing class in the holdout set, every pixel of this class in the holdout set was correctly predicted and no other classes were incorrectly predicted to belong to the aquatic floating class.", "format": "SHP", "hash": "", "id": "c2dacb03-e8bf-48cb-bccf-45b3661a6b73", "last_modified": "2023-05-17T01:30:55.346379", "metadata_modified": "2025-12-20T04:33:22.529063", "mimetype": "application/zip", "mimetype_inner": null, "name": "Bool and Hacks Lagoons Landcover Model Output 2022.zip", "package_id": "e99ccbc7-ff9f-4ccc-b20e-ebaee2239189", "position": 0, "resource_type": null, "size": 2266752, "state": "active", "url": "https://data.sa.gov.au/data/dataset/e99ccbc7-ff9f-4ccc-b20e-ebaee2239189/resource/c2dacb03-e8bf-48cb-bccf-45b3661a6b73/download/bool-and-hacks-lagoons-landcover-model-output-2022.zip", "url_type": null, "zip_extract": false}], "tags": [{"display_name": "2022", "id": "d03c4266-840b-4b28-85c0-1f14fa7d9a85", "name": "2022", "state": "active", "vocabulary_id": null}, {"display_name": "Bool Lagoon Game Reserve", "id": "bafc1dd3-4bfe-4a51-9b1d-f127d86cc148", "name": "Bool Lagoon Game Reserve", "state": "active", "vocabulary_id": null}, {"display_name": "Hacks Lagoon Conservation Park", "id": "3a8ef0fd-29e5-4087-a33a-ef830fb2c3c8", "name": "Hacks Lagoon Conservation Park", "state": "active", "vocabulary_id": null}, {"display_name": "Landcover Model", "id": "f7ca8e0c-c5c4-4250-813d-76f16a2b7892", "name": "Landcover Model", "state": "active", "vocabulary_id": null}, {"display_name": "Limestone Coast", "id": "0b4bef73-3d44-4479-b8f5-b65040282308", "name": "Limestone Coast", "state": "active", "vocabulary_id": null}, {"display_name": "inundation", "id": "5dc57b60-08b4-4767-b902-28db25ea0045", "name": "inundation", "state": "active", "vocabulary_id": null}, {"display_name": "vegetation", "id": "b0597888-1bf5-41a5-85c1-d547d3ef3982", "name": "vegetation", "state": "active", "vocabulary_id": null}, {"display_name": "wetland", "id": "4aa25504-6277-4486-b7e5-8305c012c096", "name": "wetland", "state": "active", "vocabulary_id": null}], "groups": [], "relationships_as_subject": [], "relationships_as_object": []}}