ESC-50 is a curated dataset of 2,000 environmental audio recordings organized into 50 sound classes, such as animal sounds, natural noises, and human activities. Each audio clip is five seconds long and manually labeled to ensure high-quality annotations. Developed by Karol J. Piczak, the dataset is designed to provide a standardized benchmark for environmental sound classification.
© 2026 - Copyright AIKosh. All rights reserved.