Visual recognition of guitar chords using neural networks

Abstract

In this paper, we use deep learning to study high-level features for guitar chord detection. In particular, the goal of this project is to build a neural network capable of recognizing finger patterns on the frets of a guitar. Given an input image, the network is able to identify fingers, frets, strings and the corresponding chord. Using a 2-stack Hourglass network for the detection and applying a post-processing algorithm to its corresponding output heatmaps, a 97% accuracy on the detection of chords of 205 different images is obtained