stevetjoa

I've been working on Violin.io (https://violin.io) since 2015. It's eventually intended to be interactive sheet music in the browser synced with YouTube recordings of classical music through the use of music information retrieval (MIR) algorithms. For example, click on the score at measure 50, and you'll get a YouTube video at measure 50, and you can change among different performances of that work at measure 50.

To view my rejected application for the inaugural Classical Goes Tech pitch competition last November, see https://stevetjoa.com/classicalgoestech/.

There remains technical challenges, so for now you won't find much yet other than well-tagged and categorized videos of classical works without sheet music. But I'm actively working on it.

You can help. If you frequently listen to or perform classical music, contact me at steve at violin io. I would like to know what specific works interest you. Currently, I'd prefer to focus on strings or piano, and Bach/Mozart/Beethoven/Brahms, and preferably chamber. But I'm open to all suggestions. (I think there's merit to pleasing a narrow audience very well, and with classical music variety being enormous, it can get out of hand quickly.)

Sometimes I cope by playing with my eyes closed. Violinists have options where to position their eyes: on their bow, on their fingers (rare, but helpful sometimes in practice), out into space, or closed. For example, James Ehnes and Hilary Hahn look at their bows, but Itzhak Perlman and many others play with their eyes closed. I focus on the bow, but when someone is taking a video, I might close my eyes to remove the distractions.

And yes, you're also correct that for some instruments, concerts are not easy to come by. Our orchestra has a wait list of 70 flutes! Our #1 organizational challenge is figuring out how to provide more performance opportunities to more people in a financially sustainable way.

(Incidentally, I have seen one solo flautist busking in Powell Station. He was a high school student.)

My focus on software is also 0-5 hours a day, and those hours are only before noon and after 10 pm.

Consequently, another hot research area is image anti-forensics -- the obfuscation of such operations to avoid detection. (Shameless self-promotion: one example paper on anti-forensics of JPEG compression can be found here [2].)

[1] http://www.signalprocessingsociety.org/publications/periodic...

[2] http://www.mirlab.org/conference_papers/International_Confer...

I would gladly welcome if you or someone could write a guide that has the comprehensiveness of the PDF above but with more NLP domain-specific discussion and concrete examples.

Yeah, positioning is key. I do not disrupt traffic flow at all; for those who know Powell, I stand at the end of the long hallway on the 4th street side. That gets me a captive audience for about 30 yards. They have no choice but to listen to me on their way to the BART/Muni gates. Acoustics are good too.

Indoors, there is little chance of bills being blown away, but otherwise, I agree with you about coins. $1 coins are actually not that uncommon.

You learn a lot about people and crowd dynamics. Donations come in clusters; if a person sees someone else donate, he/she is more likely to. In an attempt to teach empathy, mothers often give cash to their three-year old kid to put in the violin case. The few who drop a $10 or $20 are often over 40 years old.

Twice I've had money stolen, both by dudes in their late teens or early twenties.

Sound definitely matters; you can't just "look good". Busking at the Ferry Building was a bad idea; no one can hear you. I'll never do that again.

High-density foot traffic is not always a good thing. In the evening rush hour, there are so many people obstructing view and making noise that I imagine it's difficult to appreciate the music by the time you've reached the fare gates. I believe that a medium traffic density works best where the sound can travel and there is always a direct line of sight between pedestrian and performer.

Sometimes the "great works" make poor busking choices. The Bach Chaconne is a great example.

Unsurprisingly, an ensemble with more performers earns more than a soloist. On August 20, 2014, we played the Mendelssohn Octet in Powell Station and collectively made $461.64 over 90 minutes, or $307.76 per hour. However, that's $38.47 per hour per person. I make $15-50 per hour solo. So I wonder where the diminishing returns begin, i.e. what ensemble size maximizes revenue per hour per person.

Is it scary? The scariest part was leaving the apartment that first day. Exiting the apartment door was the point of no return. "Okay, I have my jacket and shoes on and my violin, I might as well go on with it." That moment when, for the first time, alone, I plopped my music stand in the middle of Embarcadero Station was one of the most uncomfortable moments of my life. However, once you start playing, that's easy. The music is the easy part.

If you want to busk but are scared, I recommend going at a less busy time, e.g. Saturday morning at 10 am. Get used to the surroundings; embrace your anxiety. Once you're comfortable, try again next week at 2 pm.

As I discovered just now, the whole thing is available on YouTube [https://www.youtube.com/watch?v=TxEcifeMf08, only 1.5 hours, not 3 hours as the video incorrectly suggests].

A collection of instructional IPython notebooks related to music information retrieval.

It's clearly incomplete, but perhaps it might help someone. Pull requests are welcome.

[1] https://squareup.com/help/us/en/article/5372-square-cash-wit...

http://nbviewer.ipython.org/github/stevetjoa/stanford-mir/bl...

https://github.com/stevetjoa/stanford-mir/blob/master/notebo...

Recent HN discussion on another article about orchestral conductors: https://news.ycombinator.com/item?id=7687641

Chapter 1: http://onlinestatbook.com/2/introduction/introduction.html

http://www.seattlepi.com/news/world/article/American-missing...

http://www.wjla.com/articles/2015/01/hari-simran-sterling-ma...

We began using these as instructional material at a summer workshop on MIR at CCRMA, Stanford [https://ccrma.stanford.edu/workshops/music-information-retri...]. For many years, we used Matlab, but we decided it was finally time to move over to Python. The GitHub repo includes a Vagrantfile that allows you to use Vagrant to provision a Virtualbox VM with all the audio libraries preloaded.

For included technologies, see the Vagrant box website: https://vagrantcloud.com/stevetjoa/boxes/stanford-mir/versio...

Your feedback, including pull requests and issues (and stars!), are most welcome.

Relevant plug: I have a Vagrant box [1] and GitHub repo [2] containing IPython notebooks that we use in a workshop on music information retrieval. (Caution: the IPython notebooks are under heavy development, i.e. incomplete. If you have any feedback, please create a GitHub issue.) I just added pydub to the latest version of the Vagrant box [3].

The nice thing about this setup is that, regardless of your host OS, everybody has the same development environment so you don't have to go through the pains of installing numpy, scipy, scikit-learn, essentia, and more.

[1] https://vagrantcloud.com/stevetjoa/boxes/stanford-mir

[2] https://github.com/stevetjoa/stanford-mir

[3] https://vagrantcloud.com/stevetjoa/boxes/stanford-mir/versio...

In short, LSH is an algorithm that hashes points that are nearby in a feature space into the same bin with high probability. Contrast that with cryptographically secure hashes where the tiniest change in the input is designed to yield a completely different hash. The point is that, in domains like multimedia, you want to tolerate some distortions to your signal, e.g. microphone noise, blur, etc. These minor distortions shouldn't affect your characterization of the data, e.g. "is this a guitar", "is this a cat", etc.

The advantages are that it's simple to implement, and it has mathematically provable probability bounds and query complexity.

[1] http://stackoverflow.com/questions/5751114/nearest-neighbors...

[2] http://www.cs.princeton.edu/courses/archive/spr05/cos598E/bi...

[3] http://www.vldb.org/conf/1998/p194.pdf

[4] http://www.vldb.org/conf/1999/P49.pdf

[5] http://web.iitd.ac.in/~sumeet/Slaney2008-LSHTutorial.pdf

A good strategy is to wash towels and pants yourself (these are heavier and therefore costlier to wash, and relatively easy to fold and put away yourself), and give the shirts and jackets to wash.io to wash, press, and fold.

It seems like the AWS JS SDK is the youngest, least mature (there are surprisingly few results on Google), and perhaps most difficult to use, but it's sure nice that I can manage my existing AWS stuff with a client-side JS SDK, now. If you use AWS JS SDK with DynamoDB, isn't it as if you've built your own Firebase?

Someone has to make the musical decisions. In a string quartet, it might be the first violin. In a baroque ensemble, as the article stated, it might be the harpsichordist. In a symphony, it's the conductor.

Sure, a conductor keeps time, but not so much during the performance itself. With enough practice, any professional orchestra should be able to keep time on its own, but during rehearsal, someone still needs to say exactly what that tempo is in the first place.

Having played with many amateur orchestras, currently rehearsing with one weekly, I've seen good and bad conductors. Just like a CEO or basketball coach or political leader, the conductor shares his/her vision of "success" yet knows the low-level technical details of the craft required to lead them there. The part about waving one's arms on stage has little to do with it.

An organization -- any organization -- acquires the characteristics of its leader. Therefore, one must not underestimate the importance of a good conductor.