EG-IPT Dataset (Electric Guitar Instrumental Playing Techniques)

The Electric Guitar Instrumental Playing Techniques (EG-IPT) dataset is a large-scale audio dataset designed for the study and recognition of instrumental playing techniques (IPTs) on the electric guitar. It contains 52,320 monophonic audio files, totaling 28 hours, 22 minutes, and 56 seconds of high-resolution recordings (96kHz / 24-bit), and amounts to 29.77 GB of data.

The dataset covers 19 distinct electric guitar techniques, captured through isolated single-note performances by a professional guitarist. Recordings were conducted in a specialized studio by a sound engineer using high-quality gear, ensuring both sonic fidelity and consistency across sessions. EG-IPT offers a rich and diverse resource for research in audio analysis, IPT recognition, and machine learning.

Recording Setup

The recordings were made using a 2005 Gibson SG Standard electric guitar equipped with two humbucker pickups(bridge, neck, and combined positions). The signal was amplified through an EVH 5150 III 50W 6L6 tube amp paired with a Mesa 4×12” cabinet loaded with Celestion V30 UK speakers. All amp controls were kept flat, with no reverb applied. The sessions took place in a treated studio room (4.5m × 4.5m × 3m) and were tracked using Pro Tools Ultimate.

Microphones and Signal Chain

The guitar cabinet and environment were captured using six simultaneous microphone signals, representing specific close and ambient mic techniques: