Create OpenWhistle-CNN.json

OpenWhistle-CNN.json

@@ -0,0 +1,593 @@
+{
+  "@context": {
+    "@language": "en",
+    "@vocab": "https://schema.org/",
+    "arrayShape": "cr:arrayShape",
+    "citeAs": "cr:citeAs",
+    "column": "cr:column",
+    "conformsTo": "dct:conformsTo",
+    "containedIn": "cr:containedIn",
+    "cr": "http://mlcommons.org/croissant/",
+    "data": {
+      "@id": "cr:data",
+      "@type": "@json"
+    },
+    "dataType": {
+      "@id": "cr:dataType",
+      "@type": "@vocab"
+    },
+    "dct": "http://purl.org/dc/terms/",
+    "equivalentProperty": "cr:equivalentProperty",
+    "examples": "cr:examples",
+    "extract": "cr:extract",
+    "field": "cr:field",
+    "fileObject": "cr:fileObject",
+    "fileProperty": "cr:fileProperty",
+    "fileSet": "cr:fileSet",
+    "format": "cr:format",
+    "includes": "cr:includes",
+    "isArray": "cr:isArray",
+    "isLiveDataset": "cr:isLiveDataset",
+    "jsonPath": "cr:jsonPath",
+    "key": "cr:key",
+    "md5": "cr:md5",
+    "parentField": "cr:parentField",
+    "path": "cr:path",
+    "prov": "http://www.w3.org/ns/prov#",
+    "rai": "http://mlcommons.org/croissant/RAI/",
+    "recordSet": "cr:recordSet",
+    "references": "cr:references",
+    "regex": "cr:regex",
+    "repeated": "cr:repeated",
+    "replace": "cr:replace",
+    "samplingRate": "cr:samplingRate",
+    "sc": "https://schema.org/",
+    "separator": "cr:separator",
+    "source": "cr:source",
+    "subField": "cr:subField",
+    "transform": "cr:transform"
+  },
+  "@type": "sc:Dataset",
+  "conformsTo": [
+    "http://mlcommons.org/croissant/1.1",
+    "http://mlcommons.org/croissant/RAI/1.0"
+  ],
+  "name": "OpenWhistle-CNN",
+  "alternateName": [
+    "OpenWhistleNeurIPS26/OpenWhistle-CNN",
+    "OpenWhistle-1.0-CNN"
+  ],
+  "description": "OpenWhistle 1.0 CNN is a public binary dolphin whistle detection dataset for CNN training and evaluation. It contains 76,516 audio windows with rendered spectrogram images, binary labels, filenames, recording identifiers, and onset/offset timestamps. The full dataset has balanced noise and whistle classes across train, validation, and test splits; a deterministic 480-example review-sample configuration is provided for quick inspection.",
+  "url": "https://huggingface.co/datasets/OpenWhistleNeurIPS26/OpenWhistle-CNN",
+  "citeAs": "OpenWhistle: A Large-Scale Longitudinal Dataset and Benchmark of Bottlenose Dolphin Vocalizations. NeurIPS 2026 anonymous submission.",
+  "creator": {
+    "@type": "Organization",
+    "name": "Anonymous NeurIPS 2026 authors",
+    "url": "https://huggingface.co/OpenWhistleNeurIPS26"
+  },
+  "publisher": {
+    "@type": "Organization",
+    "name": "OpenWhistleNeurIPS26",
+    "url": "https://huggingface.co/OpenWhistleNeurIPS26"
+  },
+  "datePublished": "2026-04-06",
+  "dateModified": "2026-04-27",
+  "license": "https://creativecommons.org/licenses/by/4.0/",
+  "version": "1.0.0",
+  "identifier": "https://huggingface.co/datasets/OpenWhistleNeurIPS26/OpenWhistle-CNN",
+  "keywords": [
+    "dolphin",
+    "bioacoustics",
+    "whistle-detection",
+    "audio",
+    "spectrogram",
+    "cnn",
+    "binary classification",
+    "passive acoustic monitoring",
+    "Hugging Face",
+    "Croissant",
+    "Responsible AI"
+  ],
+  "isAccessibleForFree": true,
+  "isLiveDataset": false,
+  "spatialCoverage": {
+    "@type": "Place",
+    "name": "Dolphin Reef, Eilat, northern Gulf of Aqaba"
+  },
+  "distribution": [
+    {
+      "@type": "cr:FileObject",
+      "@id": "repo",
+      "name": "repo",
+      "description": "The Hugging Face git repository containing the dataset parquet conversion.",
+      "contentUrl": "https://huggingface.co/datasets/OpenWhistleNeurIPS26/OpenWhistle-CNN/tree/refs%2Fconvert%2Fparquet",
+      "encodingFormat": "git+https",
+      "sha256": "https://github.com/mlcommons/croissant/issues/80"
+    },
+    {
+      "@type": "cr:FileSet",
+      "@id": "parquet-files-for-config-default",
+      "containedIn": {
+        "@id": "repo"
+      },
+      "encodingFormat": "application/x-parquet",
+      "includes": "default/*/*.parquet"
+    },
+    {
+      "@type": "cr:FileSet",
+      "@id": "parquet-files-for-config-review-sample",
+      "containedIn": {
+        "@id": "repo"
+      },
+      "encodingFormat": "application/x-parquet",
+      "includes": "review-sample/*/*.parquet"
+    }
+  ],
+  "recordSet": [
+    {
+      "@type": "cr:RecordSet",
+      "@id": "default_splits",
+      "name": "default_splits",
+      "description": "Splits for the default config.",
+      "dataType": "cr:Split",
+      "key": {
+        "@id": "default_splits/split_name"
+      },
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "default_splits/split_name",
+          "dataType": "sc:Text"
+        }
+      ],
+      "data": [
+        {
+          "default_splits/split_name": "test"
+        },
+        {
+          "default_splits/split_name": "train"
+        },
+        {
+          "default_splits/split_name": "validation"
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "default",
+      "name": "default",
+      "description": "Full binary CNN whistle detection configuration with balanced noise and whistle classes across train, validation, and test splits.",
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "default/split",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "fileProperty": "fullpath"
+            },
+            "transform": {
+              "regex": "default/(?:partial-)?(test|train|validation)/.+parquet$"
+            }
+          },
+          "references": {
+            "field": {
+              "@id": "default_splits/split_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/audio",
+          "dataType": "sc:AudioObject",
+          "samplingRate": 96000,
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "audio"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/spectrogram",
+          "dataType": "sc:ImageObject",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "spectrogram"
+            },
+            "transform": {
+              "jsonPath": "bytes"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/label",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "label"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/file_name",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "file_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/recording",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "recording"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/onset",
+          "dataType": "cr:Float64",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "onset"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "default/offset",
+          "dataType": "cr:Float64",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-default"
+            },
+            "extract": {
+              "column": "offset"
+            }
+          }
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "review-sample_splits",
+      "name": "review-sample_splits",
+      "description": "Splits for the review-sample config.",
+      "dataType": "cr:Split",
+      "key": {
+        "@id": "review-sample_splits/split_name"
+      },
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample_splits/split_name",
+          "dataType": "sc:Text"
+        }
+      ],
+      "data": [
+        {
+          "review-sample_splits/split_name": "train"
+        },
+        {
+          "review-sample_splits/split_name": "test"
+        }
+      ]
+    },
+    {
+      "@type": "cr:RecordSet",
+      "@id": "review-sample",
+      "name": "review-sample",
+      "description": "Deterministic 480-example reviewer inspection subset preserving train/test separation and balancing noise and whistle labels within each reviewer split.",
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/split",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "fileProperty": "fullpath"
+            },
+            "transform": {
+              "regex": "review\\-sample/(?:partial-)?(train|test)/.+parquet$"
+            }
+          },
+          "references": {
+            "field": {
+              "@id": "review-sample_splits/split_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/audio",
+          "dataType": "sc:AudioObject",
+          "samplingRate": 96000,
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "audio"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/spectrogram",
+          "dataType": "sc:ImageObject",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "spectrogram"
+            },
+            "transform": {
+              "jsonPath": "bytes"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/label",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "label"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/file_name",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "file_name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/recording",
+          "dataType": "sc:Text",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "recording"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/onset",
+          "dataType": "cr:Float64",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "onset"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "review-sample/offset",
+          "dataType": "cr:Float64",
+          "source": {
+            "fileSet": {
+              "@id": "parquet-files-for-config-review-sample"
+            },
+            "extract": {
+              "column": "offset"
+            }
+          }
+        }
+      ]
+    }
+  ],
+  "rai:dataCollection": "The dataset was assembled from OpenWhistle passive acoustic monitoring recordings at Dolphin Reef, Eilat, a semi-natural marine site on the northern Gulf of Aqaba. Fixed hydrophones recorded a stable pod of known dolphins. The released dataset contains short audio windows, rendered spectrogram images, binary whistle/noise labels, recording identifiers, and onset/offset timestamps.",
+  "rai:dataCollectionType": [
+    "Physical data collection",
+    "Direct measurement",
+    "Manual Human Curator",
+    "Software Collection"
+  ],
+  "rai:dataCollectionRawData": "Raw data consisted of underwater acoustic recordings captured by fixed and hidden hydrophones. The released CNN dataset contains audio windows and spectrogram images with public columns: audio, spectrogram, label, file_name, recording, onset, and offset. Labels are binary: noise=0 and whistle=1.",
+  "rai:dataCollectionMissingData": "The public dataset does not include per-example video context, exact hydrophone coordinates, environmental conditions, full behavioral context, whistle-type labels as a primary task label, or human-identifying information. The dataset is window-level and does not provide complete continuous-recording annotation beyond the released windows.",
+  "rai:dataPreprocessingProtocol": [
+    "Audio recordings were converted into short binary-labeled windows and rendered spectrogram images for CNN training.",
+    "The full dataset contains 76,516 rows: 53,828 train, 5,980 validation, and 16,708 test examples. Each split is exactly balanced between noise and whistle labels.",
+    "Train and validation examples come from the non-2019/2020 pool, while the test split is a manual 2019-2020 test split built from the full classification all configuration.",
+    "The review-sample configuration was generated deterministically with seed 42, preserving train/test separation and balancing noise and whistle examples within each reviewer split."
+  ],
+  "rai:dataAnnotationProtocol": "Each item has a binary label indicating noise or whistle. Whistle windows are derived from manually curated whistle annotations, while noise windows were selected as non-whistle examples. The dataset is intended for binary whistle presence detection rather than fine-grained whistle-type classification.",
+  "rai:annotationsPerItem": "Each item has one binary class label: noise or whistle.",
+  "rai:machineAnnotationTools": [
+    "Spectrogram rendering software was used to create the spectrogram image column from audio windows.",
+    "Hugging Face Datasets and parquet conversion tooling were used to package and publish the dataset."
+  ],
+  "rai:dataUseCases": [
+    "Construct represented: short underwater acoustic windows and spectrograms labeled for binary dolphin whistle presence or non-whistle noise.",
+    "Validated use case: training, validation, and testing of CNN-based binary dolphin whistle detectors.",
+    "Validated use case: benchmarking audio-window and spectrogram-image approaches for whistle-vs-noise classification.",
+    "Validated use case: training the OpenWhistle VGG16-style CNN detector used in the OpenWhistle processing pipeline.",
+    "Supported reviewer use case: manual inspection through the deterministic review-sample configuration.",
+    "Not validated use case: full behavioral interpretation, individual identification, whistle-type classification, or continuous-recording segmentation without additional post-processing and local validation."
+  ],
+  "rai:dataLimitations": [
+    "The dataset is designed for binary window-level whistle-vs-noise classification, not full behavioral interpretation, individual identification, whistle-type classification, or continuous-recording segmentation without additional post-processing.",
+    "The full split design uses a manual 2019-2020 test split and non-2019/2020 train/validation pool, so performance may reflect temporal and recording-condition differences between these periods.",
+    "The dataset comes from a semi-natural site with a stable pod of known dolphins, so model performance may not generalize to fully wild populations, other species, other recording devices, or different acoustic environments without validation.",
+    "Exact binary balance improves classifier training and comparison but does not reflect natural whistle/noise prevalence in continuous acoustic recordings.",
+    "The review-sample configuration is for inspection only and should not replace the full dataset for model development or reporting."
+  ],
+  "rai:dataBiases": [
+    "The dataset reflects Dolphin Reef recordings and may overrepresent acoustic, social, and environmental conditions at that semi-natural site.",
+    "The dataset is intentionally class-balanced, which differs from natural passive acoustic recordings where noise/non-whistle windows may dominate.",
+    "Spectrogram rendering choices and CNN-oriented preprocessing may bias models toward image-like features and may not transfer directly to raw-audio models without validation.",
+    "Curated positive and negative window selection may underrepresent ambiguous, overlapping, low-SNR, or rare acoustic events.",
+    "The manual 2019-2020 test split may emphasize conditions specific to those recording years."
+  ],
+  "rai:personalSensitiveInformation": [
+    "No human subjects are involved in the dataset. Public columns contain animal audio windows, spectrogram images, labels, recording identifiers, filenames, and timing metadata.",
+    "The data consists of underwater animal acoustic recordings. Users should nevertheless treat incidental non-target sounds as possible and avoid attempts to infer people, vessels, precise locations, or sensitive field-site details from audio, filenames, or recording identifiers."
+  ],
+  "rai:dataSocialImpact": "The dataset supports open research on passive acoustic monitoring, dolphin bioacoustics, and non-invasive whistle detection. Positive impacts include reproducible training and evaluation of CNN whistle detectors and improved tooling for animal communication research. Misuse risks include deploying detectors in new sites without local validation, treating window-level detection as behavioral or ecological evidence, or ignoring failures under acoustic conditions not represented in the curated windows. The original data collection was passive and non-invasive, did not interfere with dolphin behavior, and involved no human subjects.",
+  "rai:hasSyntheticData": false,
+  "rai:dataReleaseMaintenancePlan": "The dataset is hosted publicly on Hugging Face under OpenWhistleNeurIPS26/OpenWhistle-CNN and released under CC-BY 4.0. The repository includes deterministic review-sample data for reviewer inspection. Future OpenWhistle releases may update CNN training data, split documentation, labels, or evaluation protocols.",
+  "prov:wasDerivedFrom": [
+    {
+      "@id": "urn:openwhistle:cnn-source-recordings-and-detection-windows:anonymous-neurips-2026",
+      "@type": "prov:Entity",
+      "name": "Anonymous OpenWhistle passive acoustic recordings and curated CNN detection windows",
+      "description": "Underlying passive acoustic recordings and curated binary detection windows used to derive the public CNN dataset. The source is described anonymously during double-blind review.",
+      "license": "https://creativecommons.org/licenses/by/4.0/",
+      "prov:label": "OpenWhistle passive acoustic recordings and curated CNN detection windows",
+      "sc:license": "https://creativecommons.org/licenses/by/4.0/",
+      "prov:wasAttributedTo": {
+        "@id": "anonymous_openwhistle_authors",
+        "name": "Anonymous OpenWhistle authors",
+        "prov:label": "Anonymous OpenWhistle authors"
+      }
+    }
+  ],
+  "prov:wasGeneratedBy": [
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q4929239"
+      },
+      "name": "Passive acoustic data collection",
+      "description": "Underwater audio was collected passively using fixed and hidden hydrophones. The collection did not interfere with dolphin behavior, did not train or constrain animals, and involved no human subjects.",
+      "prov:label": "Passive acoustic data collection",
+      "sc:description": "Underwater audio was collected passively using fixed and hidden hydrophones. The collection did not interfere with dolphin behavior, did not train or constrain animals, and involved no human subjects.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:Agent",
+          "@id": "anonymous_openwhistle_authors",
+          "name": "Anonymous OpenWhistle authors",
+          "description": "Anonymous research team responsible for the OpenWhistle data release and documentation during double-blind review.",
+          "prov:label": "Anonymous OpenWhistle authors",
+          "sc:description": "Anonymous research team responsible for the OpenWhistle data release and documentation during double-blind review."
+        }
+      ]
+    },
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q109719325"
+      },
+      "name": "Binary window labeling",
+      "description": "Audio windows were labeled as noise or whistle. Whistle windows are derived from curated whistle annotations, and noise windows provide non-whistle examples for binary detection.",
+      "prov:label": "Binary window labeling",
+      "sc:description": "Audio windows were labeled as noise or whistle. Whistle windows are derived from curated whistle annotations, and noise windows provide non-whistle examples for binary detection.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:Agent",
+          "@id": "anonymous_openwhistle_authors",
+          "name": "Anonymous OpenWhistle authors",
+          "description": "Anonymous research team responsible for curation and binary label mapping during double-blind review.",
+          "prov:label": "Anonymous OpenWhistle authors",
+          "sc:description": "Anonymous research team responsible for curation and binary label mapping during double-blind review."
+        }
+      ]
+    },
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q5227332"
+      },
+      "name": "Spectrogram rendering and dataset packaging",
+      "description": "Audio windows were converted into rendered spectrogram images and packaged with audio, labels, filenames, recording identifiers, and onset/offset metadata as Hugging Face parquet configurations.",
+      "prov:label": "Spectrogram rendering and dataset packaging",
+      "sc:description": "Audio windows were converted into rendered spectrogram images and packaged with audio, labels, filenames, recording identifiers, and onset/offset metadata as Hugging Face parquet configurations.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:SoftwareAgent",
+          "@id": "spectrogram_rendering_pipeline",
+          "name": "Spectrogram rendering pipeline",
+          "description": "Software pipeline used to create spectrogram images from audio windows.",
+          "prov:label": "Spectrogram rendering pipeline",
+          "sc:description": "Software pipeline used to create spectrogram images from audio windows."
+        },
+        {
+          "@type": "prov:SoftwareAgent",
+          "@id": "hugging_face_datasets",
+          "name": "Hugging Face Datasets",
+          "description": "Software tooling used to package and publish the dataset with Audio and Image features.",
+          "prov:label": "Hugging Face Datasets",
+          "sc:description": "Software tooling used to package and publish the dataset with Audio and Image features."
+        },
+        {
+          "@type": "prov:Agent",
+          "@id": "anonymous_openwhistle_authors",
+          "name": "Anonymous OpenWhistle authors",
+          "description": "Anonymous research team responsible for preprocessing, curation, and release during double-blind review.",
+          "prov:label": "Anonymous OpenWhistle authors",
+          "sc:description": "Anonymous research team responsible for preprocessing, curation, and release during double-blind review."
+        }
+      ]
+    },
+    {
+      "@type": "prov:Activity",
+      "prov:type": {
+        "@id": "https://www.wikidata.org/wiki/Q3306762"
+      },
+      "name": "Review sample generation",
+      "description": "A deterministic 480-example review-sample configuration was generated with seed 42. It preserves train/test separation and balances noise and whistle examples within each reviewer split.",
+      "prov:label": "Review sample generation",
+      "sc:description": "A deterministic 480-example review-sample configuration was generated with seed 42. It preserves train/test separation and balances noise and whistle examples within each reviewer split.",
+      "prov:wasAttributedTo": [
+        {
+          "@type": "prov:Agent",
+          "@id": "anonymous_openwhistle_authors",
+          "name": "Anonymous OpenWhistle authors",
+          "description": "Anonymous research team responsible for creating reviewer inspection subsets.",
+          "prov:label": "Anonymous OpenWhistle authors",
+          "sc:description": "Anonymous research team responsible for creating reviewer inspection subsets."
+        }
+      ]
+    }
+  ]
+}