Field cricket (Plebeiogryllus guttiventris) acoustic dataset for machine learning

This archive provides the complete dataset used in both the study by Diptarup and Rohini (2013), "Call intensity is a repeatable and dominant acoustic featuredetermining male call attractiveness in a field cricket", and our study, "Acoustic individual identification in a species of field cricket using deep learning". The archive contains raw audio recordings and their associated temperature metadata. Files provided The raw_audio_files.zip zipped file contains 138 raw audio files . The accompanying recording_temperature_data.xlsx spreadsheet provides temperature metadata associated with each recording. The audio_naming_convention.docx file describes the recording naming scheme, allowing users to identify the individual cricket from which a recording was obtained, the recording night, and the order of the recording within that night. The train_validation_test_datasets.zip zipped file contains the train, validation, and test datasets used in our experiments for both closed- and open-population individual identification. The open-population setting, includes scenarios involving both individual seen during training and novel individuals only appearing in the test sets. For each identificatiton setting, datasets are provided using both 1-second segments and 5-syllable segments, thereby offering two complementary temporal scales. Each dataset .csv file contains: The segment name; The segment ID, i.e., the source individual; The segment start and end times within the source recording; The recording night; The order of the recording within that night; The name of the source audio file from which the segment was extracted; These metadata are essential for generating 1-second and 5-syllable audio segments and their corresponding labels from the raw recordings, as well as for constructing within-night and across-night evaluation datasets. Implementation details Detailed information on train-validation-test set construction, audio preprocessing, feature extraction, model development, training and evaluation procedures is available in the accompanying software archive: DOI: 10.5281/zenodo.20542968 Users interested in reproducing experiments reported in the "Acoustic individual identification in a species of field cricket using deep learning" paper should consult the software archive in conjuction with this dataset. Citation Users of this dataset should cite both the dataset itself and the associated publications listed above.