Finding the genre of a song with Deep Learning

[iverjo]

To the author: Have you tried to use a logarithmic frequency scale in the spectrogram? [1] That representation is closer to the way humans perceive sound, and gives you finer resolution in the lower frequencies. [2] If you want to make your representation even closer to the human's perception, take a look at Google's CARFAC research. [3] Basically, they model the ear. I've prepared a Python utility for converting sound to Neural Activity Pattern (resembles a spectrogram when you plot it) here: https://github.com/iver56/carfac/tree/master/util

[1] https://sourceforge.net/p/sox/feature-requests/176/

[2] https://en.wikipedia.org/wiki/Mel_scale

[3] http://research.google.com/pubs/pub37215.html

[stakecounter]

Mel scale spectrograms are the approach taken in a research paper which uses roughly the same technique as is described in this post: https://dl.dropboxusercontent.com/u/19706734/paper_pt.pdf

[Terribledactyl]

I don't think this problem is bound by absolute frequency resolution, the tightest distance between two notes on a typical piano is ~2hz and if you assume a doubling between octaves you're at <90 notes. The temporal changes and relative chord progressions probably give more info.

[Despoisj]

I didn't intend to go that far in the human genre recognition parallel, but thanks for the references ! Good job on the script too

[nkurz]

Wow, I find it incredible that this works. As I understand it, the approach is to do a Fourier transform on a couple seconds of the song to create a 128x128 pixel spectrogram. Each horizontal pixel represents a 20 ms slice in time, and each vertical pixel represents 1/128 of the frequency domain.

Then treating these spectrograms as images, train a neural net to classify them using pre-labelled samples. Then take samples from the unknown songs, and let it classify them. I find it incredible that 2.5 seconds of sound represented as a tiny picture captures information enough for reliable classification, but apparently it does!

[iammyIP]

One reason might be that the mentioned genres are highly formulaic to begin with. The standard rap song contains about 2 bars of unique music stretched out over 3 minutes with slight variations. Same with dubstep and techno. All highly repetitive. Classical music got no drums, so you can detect that. Metal got guitar distortion all over the spectrum. So with these examples the spectral images should have enough distinctive features that can be learned. Why should it be different than with 'normal' pictures? Also it looks like they take four 128x128 guesses per song.

[kimburgess]

From the description in the walkthrough, it doesn't. The final output looks to be based on 5 of these slices, with each providing a probability distribution that ultimately influences the final classification.

[amelius]

I guess the spectogram behaves like an image in that translation of any feature by an arbitrary distance (dx,dy) preserves its predicting quality.

But please correct me if I'm wrong.

[Despoisj]

Yep you got it right, except the voting system adds tons of reliability because we cannot trust the slice classification (2.5s) too much.

[chestervonwinch]

1. I wonder how the continuous wavelet transform would compare to the windowed Fourier transform used here. See [1] an python implementation, for example.

2. The size of frequency analysis blocks seems arbitrary. I wonder if there is a "natural" block size based on a song's tempo, say 1 bar. This would of course require a priori tempo knowledge or a run-time estimate.

[1]: https://docs.scipy.org/doc/scipy-0.15.1/reference/generated/...

[Despoisj]

The slice size is indeed quite arbitrary, and knowing the BPM would help, but isn't reliable either (various tempos, rubato for classical etc.)

[maxerickson]

See also http://everynoise.com/ which is a view into how Spotify classifies music.

The creator wrote about it here:

http://blog.echonest.com/post/52385283599/how-we-understand-...

and writes a lot about it on their blog:

http://www.furia.com/page.cgi?terms=noise&type=search

Of course those are going in the other direction, not generating the classification from the data, but it's probably one of the best data sets as far as classifying existing music.

[Despoisj]

Thanks for the awesome ressources ! :D

[stakecounter]

Unless I'm misunderstanding the validation set, I'm skeptical of the ability of this classifier to tag unlabeled tracks, given that it is only being trained and tested on tracks which are already known to belong to one of the few trained genres. I'd be curious to see the performance if you were to additionally test on tracks which are not any of (Hardcore, Dubstep, Electro, Classical, Soundtrack and Rap), with a correct prediction being no tag.

[Despoisj]

It's true that the validation set only contains genres I used for the training. I'll try this out this evening ;)

[iverjo]

Nice approach, and well explained! By the way, Niland is a startup that also does music labeling with the help of deep learning.

Demo available here: http://demo.niland.io/

For example, it can output Drum Machine: 87%, House: 88%, Female Voice: 55%, Groovy: 93%

[Despoisj]

Thanks for the kind words, I'll take a look !

[GFK_of_xmaspast]

See also Bob Sturm's work on genre classification: http://link.springer.com/article/10.1007/s10844-013-0250-y

[Despoisj]

Wow, that's an in-depth analysis of the task ! Thanks for sharing

[tunesmith]

That's pretty cool, I'd like to use something like this to tell me what genre my own songs are. It's annoying to write a song and then upload it to some service or another and have no idea what genre to pick. :-) My stuff is somewhere in the jazz-influenced singer-songwriter american piano pop realm which is a combination that works for me but it generally feels like I'm selling the song short if I have to pick only one.

[Despoisj]

Yeah that's a problem I know - I used to make some Electro/Dubstep/Trap music - and I feel people will always disagree with the genre you pick anyway.

[return0]

Good luck convincing musicians that "THAT's your genre"

[Despoisj]

Haha, I should have titled the article "How to build an internet music-genre-troll-bot"

[dkarapetyan]

Hmm, convolution is perfectly good operation to run on wave forms as well. In fact the wikipedia article (https://en.wikipedia.org/wiki/Convolution) shows the operation on functions which would correspond to time-domain wave forms. What is the point of converting everything to pictures and then using 2D convolutions when that step could have been skipped entirely?

Converting to pictures is unnecessary. It makes the processing harder. The pooling should just happen on segments of the wave form instead of the fourier transform (frequency-domain) picture spectrograms.

[highd]

The idea is that the vertical axis of the spectrogram is basically already an hierarchical set of features (in scale/frequency). Then convolutions on that is a lot like how DenseNets combine hierarchical features.

I agree it seems a little jank, but the features are pretty good - and a lot of network architectures / training techniques are most practiced in an image processing context.

[jtmarmon]

i'm not super familiar with deep learning so forgive me if i'm missing some nuance, but what's the purpose of writing/reading to/from images? seems like it would add a ton of processing time. could the CNN not just read from a 50 item array of tuples representing the data from the 20ms slice?

[Despoisj]

I'm not sure what you mean, but I have chosen to store slices on the disk so that I could still take a look at them, and not store the data only in numpy arrays. That could be optimize for a better processing time!