In this blog post I would be
discussing about Dataset...
GTZAN Genre Collection dataset was
used to perform the classification. The dataset has been taken from the popular
software framework MARSYAS. Marsyas (Music Analysis, Retrieval and Synthesis
for Audio Signals) is an open source software framework for audio processing
with specific emphasis on Music Information Retrieval applications. Marsyas has
been used for a variety of projects in both academia and industry.
We have compared several open music
Dataset with associated metadata and select GTZAN Genre Collection, of which
contains 1000 audio tracks each 30 seconds long.
There are 10 genres represented, each
containing 100 tracks. All the tracks are 22050 Hz Mono 16 bit audio files in
.au format. The 10 music genre includes: classical, jazz, metal, pop, country,
blues, disco, metal, rock, reggae and hip-hop.
The audio files are
divided into 2 sec long audio chunks and labels are maintained accordingly.We
finally had 14000 audio samples equally distributed over 10 genres. This was
split these samples randomly into a 80-20 train-validation ratio, giving us
11200 training samples and 2800 validation samples.
Table 1. Distribution
of the Dataset
No comments:
Post a Comment