raw_data.zip Raw audio data collected in the field. It is composed of sub-folders that represent each monitoring site. Each sub-folder is composed of audio .wav files that follow the name of {site}_{date}_{time}.wav. weak_labels.csv Annotation at a 1-minute level where each raw audio data it is assigned a value representing the anuran calling activity: 0 is absence; 1 is Low; 2 is Moderate; and 3 is High. The CSV file is composed of two columns representing the site and the file name and species columns with the anuran calling activity. strong_labels.zip Annotation at a high level with temporal limits (beginning and end) of audio segments containing species-specific calls with an inter-call interval of less than 1 second. As in raw_data, each sub-folder represents a monitoring site and the files are .txt containing (i) call beginning; (ii) call end; and (iii) the species name and audio quality. anuraset.zip Preprocessed dataset with 93378 3-second audio samples input for benchmarking. The dataset folder contains 2 files and one folder containing separate folders per site. The samples are WAV audio files with fixed 3-second lengths, obtained with 22.05 kHz sampling frequency and 16-bit depth. The two other files are a README file describing the structure and construction of the dataset and a metadata CSV file containing the labels for each sample.

Abstract: Global change is predicted to induce shifts in anuran acoustic behavior, which can be studied through passive acoustic monitoring (PAM). Understanding changes in calling behavior requires the identification of anuran species, which is challenging due to the particular characteristics of neotropical soundscapes. In this paper, we introduce a large-scale multi-species dataset of anuran amphibians calls recorded by PAM, that comprises 27 hours of expert annotations for 42 different species from two Brazilian biomes. We provide open access to the dataset, including the raw recordings, experimental setup code, and a benchmark with a baseline model of the fine-grained categorization problem. Additionally, we highlight the challenges of the dataset to encourage machine learning researchers to solve the problem of anuran call identification towards conservation policy. All our experiments and resources can be found on our GitHub repository https://github.com/soundclim/anuraset.