Data for "ChickenSense: A low-cost deep learning-based solution for real-time poultry feed intake monitoring using sound technology"

The dataset includes two types of data: scale and audio. The scale data is recorded using a hanging scale and measures the feeders' weight with a 5 Hz frequency. The audio data contains 1824 hours of recorded audio using a piezo-electric sensor attached to a bucket feeder to capture vibrations as audio signals with a sampling rate of 48 kHz. Each .wav file contains 15 minutes of audio recording.
Along with the raw data, a part of the dataset containing 36 30-second audio recordings was sampled and labeled in four classes, including pecking, singing, anomaly, and silence, resulting in 3319 labeled events.