Bird Vocalization Dataset

Natural soundscapes and animal vocalizations are fascinating research subjects that provide important insights on animal behavior, populations, and ecosystems. They are investigated in the fields of ecoacoustics and bioacoustics, with a significant emphasis on signal processing and analysis. The development of more accessible digital sound recorders and significant advancements in informatics, including big data, machine learning, and signal processing, have increased computational bioacoustics in recent years. We collected a bird vocalization dataset to build a bird classier using deep learning. The dataset consists of audio recordings collected from Intaka Island, Cape Town, South Africa, for training a convolutional neural network (CNN)-based bird classifier. Recordings were gathered by placing audio recorders in various habitats across the island, for almost 5 hours.  In total, around 4 additional hours of recordings were obtained. Most recordings used a sampling rate of 48,000 Hz to capture sound details.