I now work at , a company that I founded along with 3 other genii. Tomnod applies crowdsourcing and machine learning to massive geo-spatial challenges.
We engage thousands of people online to help find the tomb of Genghis Khan, map earthquake damage in Christchurch or identify the best St. Patrick's Day parties! I'll keep this site alive for the time being but, to stay up to date with my current work, check out the Tomnod blog and join the crowd!
I'm a Ph.D
I am a founder of the UCSD Computer Audition Lab where I research:
- Machine learning models of music
- Mining information about music from waveforms, reviews, tags, web documents and more
- Online human computation for massive, distributed data collection
I am searching for the tomb of Genghis Khan with the Valley of the Khans project.
I play guitar and sing with Telefunk.Please get in touch: lukeinusagmail.com
I am very proud to report that I successfully defended my Ph.D thesis on December 2nd, 2011! The title of my dissertation is "Machines that Understand Music". My committee of Profs Gary Cottrell, Lawrence Saul, Bhaskar Rao and Nuno Vasconcelos gave me great guidance over the years and were very pleased with my defense. I'm also very grateful to the many friends who attended the talk and helped me celebrate afterwards! Of course, none of this would have been possible without the support of my advisor, Prof. Gert Lanckriet.
Onto the next adventure!
DTM Code released
I am pleased to release a Python implementation of the Dynamic Textures Mixture model that I used for music segmentation in my recent IEEE journal paper. The DTM code takes a series of audio feature vectors, merges them into longer "frames" and models each frame as a Dynamic Texture. Frames with the same Texture are assigned to the same audio segment (e.g., chours, verse, ...).
Download the code here
If you use this code in your research or if you have any questions or suggestions, please drop me a line -- I'd love to hear about it!
Journey to the Valley of the Khans
I've just returned from a month-long adventure in Mongolia, working with Dr. Albert Yu-Min Lin's Valley of the Khans project and National Geographic to search for the tomb of Genghis Khan!
We designed an built a human computation application to examine satellite images of the Mongolian steppe where Genghis lived 800 years ago. So far, over 6,000 "virtual explorers" have tagged more than a million sites of archaeological interest.
My job in Mongolia was to wake up every morning in our ger, download the new tags, identify the most interesting sites and then jump on my horse and check them out. Check out the expedition blog where I write about our discoveries.
Science in action!
"Never easy to categorise"
New Scientist featured a story
about how MIR researchers like Dan Ellis,
Luciano da F. Costa and I are making progress towards computerised music critics that can help everyone find great tunes.
The article mentions the "Bohemian Rhapsody problem" that I describe in my recent TASLP paper along with this great pic of Freddie ... perfect!
Qualcomm Innovation Fellowship
I'm proud, honored and very happy to have been awarded a Qualcomm Innovation Fellowship.
My partner in crime, Brian, and I pitched our vision of "Location-, Demographic-, Preference- and Content-based Music Discovery".
Now, with Qualcomm's help, we're actually going to make this music fan's dream a reality!
3/3 - ISMIR 2010
Congratulations to Brian McFee, Riccardo Miotto and Emanuele Coviello, three of CAL's finest who all had first-author papers accepted to ISMIR 2010!
I'm doubly happy because I was second author on all of them - check out my publications section. Utrecht, here we come!
The Science Network
A video of my talk on "Human Powered Machine Learning" has been posted on the Science Network.
I really enjoyed presenting my work on Computer Audition and Herd It to the audience at the Temporal Dynamics of Learning Center.
Finally I have confirmation of the fact that I talk too fast!
CALab undergrads win UCSD Yahoo! Hack Day
I'm very proud of Andrew Huynh and David Vanoni, two undergrads who have been working with me at CALab for the past year. Their "Rock My World" app won first prize at the Yahoo Hack Day. I had an idea for an app that would find great gigs playing in your neighborhood and, in 24hrs, they made it a reality!
Download Rock My World to your iPhone (open it in Safari and then choose "+ add to homepage").
A story about my work on building a music recommender that could be smarter than Genius by Jacobs School reporter Daniel Kane attained nerdvana by getting posted on Slashdot! This led to over 2,000 new players checking out Herd It (and a few server crashes). w00t!
Smarter Than Genius?
I just gave a talk at ISMIR in Kobe, Japan about my paper "Smarter Than Genius?", an analysis of iTunes' music recommender system. The talk was great, I had some really interesting questions and comments and it's always fun to be at ISMIR. Read the press release or download my slides
Modeling Music as a Dynamic Texture
My paper has just been accepted to IEEE Transactions on Audio, Speech and Language Processing! This paper presents a new statistical model of music, adapted from Prof. Antoni Chan's work on video, that really takes time into account. Also, it shows how to solve the Bohemian Rhapsody problem!
Listen to the 35-second version of Bohemian Rhapsody!
Herd It offical launch!
Afer years (and years) of development, testing, design, refinement, bug fixing, testing, tweaking, and playing, Herd It is fully ready to rock!
I've added a blog and a Twitter feed (@herdit) as well as a bunch of new gameplay features. Now we are making a BIG effort to promote the game - so tell your friends, newspaper, blog readers... tell everyone to play Herd It!
Once lots of people to play Herd It and we collect some great data then I can graduate!
Herd It in the news
Phenomenal reception to the Beta release of my music annotation game Herd It!
First we got written up in Voice of San Diego and that led to a news spot on Channel 6 (unfortunately, I was in Taiwan at ICASSP so Gert had to present!)
Then Robyn Williams interviewed Gert and I on the Science Show, resulting in loads of new users (at some times, too many for our server!).
Play Herd It.
ICASSP and the Bohemian Rhapsody Problem
Just gave a talk at ICASSP in Taipei, Taiwan about automatic music segmentation using dynamic texture mixture models of music.
Check out how we use this method to solve the "Bohemian Rhapsody problem" of automatically tagging muscially diverse songs!
UPDATE May 2009
An article about this research and how it contributes to our music search engine just got posted on the Jacob's School of Engineering Site, UCSD's site and EurekAlert.org!
Entertainment Tired Love New Speech!
News of my talk at the Math Club last month has appeared in the Swedish newspaper, DN.
(no, I can't understand it either. Try the humourously flawed Google translation!).
Download my slides on "Machines that Understand Music".
- Jacbos School - Electrical Engineer Explores Mongolia - And You Can Too
- CalIT2 - Citizen Scientists Explore Ancient Mongolia from Afar
- Gizmodo - Help Survey Genghis Khan's Lost Tomb With Some Armchair Archaeology
- Wired - Help Find Genghis Khanís Tomb From the Comfort of Your Home
- New Scientist - Computerised critics could find the music you'll like
- Slashdot - Going Head To Head With Genius On Playlists
- IGN.com - Herd It's New Take on Music Apps
- The Science Show - Computer trained to describe music
- ZDNet - Facebook game aids development of new 'Google for music' search engine
- Wired - Gadgets Join the Search for the Lost Tomb of Genghis Khan
- Alt Search Engines - Have you Herd It? Itís a new music search engine!
- The Atlantic Wire - 5 Predictions for the Future of Music
- The Register - Student boffins take on iTunes' not-so-smart Genius
- MIT Technology Review - Software with a Better Ear for Music
- PhysOrg.com - From a Queen song to a better music search engine
- Voice of San Diego - 'Black Box' Could Redefine the Search for Music
- MIT Technology Review - Getting Computers Into the Groove
- NBC News - Music That's Made for You
- San Diego Channel 6 - UCSD Creates a Computer that KNOWS Good Music? (TV spot here)
Luke Barrington's Publications
Barrington Turnbull & Lanckriet - Game-Powered Machine Learning.
Proceedings of the National Acadmeny of Sciences of the United States of America. (Herd It game)
To appear in IEEE Transactions on Audio, Speech and Language Processing, 2012.
Barrington, Ghosh, Greene, Har-Noy, Berger, Gill, Lin and Huyck -
Crowdsourcing earthquake damage assessment using remote sensing imagery
Annals of Geophysics 54(6) 2011
Barrington, Har-Noy, Ricklin, Stastny and Rainey -
Computer Vision methods for Ship Identification in Cloudy Satellite Imagery
IEEE Applied Imagery Pattern Recognition (AIPR) Workshop 2011
Miotto, Barrington & Lanckriet - Improving Auto-tagging by Modeling Semantic Co-occurrences.
Barrington, Oda & Lanckriet - Smarter Than Genius? Human Evaluation of Music Recommender Systems.
ISMIR 2009 (slides)
Barrington, Turnbull, O'Malley & Lanckriet -
User-Centered Design of a Social Game to Tag Music.
HCOMP 2009 (slides)
Barrington, Turnbull, Yazdani & Lanckriet -
Combining Audio Content and Social Context for Semantic Music Discovery.
Barrington, Marks, Hsiao & Cottrell -
NIMBLE: A Kernel Density Model of Saccade-Based Visual Memory.
Journal of Vision, 8(14):17, 1-14
Barrington, Yazdani, Turnbull & Lanckriet -
Combination of Feature Kernels for Semantic Music Retrieval.
ISMIR 2008 (poster)
Turnbull, Barrington, Torres & Lanckriet 2008 -
Semantic Annotation and Retrieval of Music and Sound Effects.
IEEE Transactions on Audio, Speech, and Language Processing, 16(2), 467-476
Torres, Turnbull, Barrington & Lanckriet 2007 -
Identifying Words that are Musically Meaningful.
International Symposium on Music Information Retrieval (ISMIR), Vienna, September 2007
Turnbull, Liu, Barrington & Lanckriet 2007 -
A Game-Based Approach for Collecting Semantic Annotations of Music.
International Symposium on Music Information Retrieval, Vienna, September 2007
Turnbull, Barrington, Torres & Lanckriet 2007 -
Towards Musical Query-by-Semantic-Description using the CAL500 Data Set.
ACM Special Interest Group on Information Retrieval, Amsterdam
Barrington, Marks & Cottrell 2007 -
NIMBLE: A Kernel Density Model of Saccade-Based Visual Memory.
Cognitive Science Society Conference, Nashville
Barrington, Chan, Turnbull & Lanckriet 2007 -
Audio Information Retrieval Using Semantic Similarity.
International Conference on Acoustics, Speech and Signal Processing, Hawai'i
Turnbull, Barrington & Lanckriet 2006 -
Modelling Music and Words.
International Symposium on Music Information Retrieval, Victoria
Barrington & Cottrell 2006 -
Automatic Visual Integration - Defragmenting the Face.
Cognitive Science Society Conference, Vancouver (slides 492kb zip file)
Temporal Dynamics of Learning Center all-hands meeting on "Human Powered Machine Learning", Jan 23rd 2010 (Slides).
music discovery engine can help musicians like my friends SO3, Juna and Beat7.
Herd It works and how it helps undiscovered artists.
Audio clip of Gert and me on the Science Show talking about Herd It .
Slides for PEN Online talk on NIMBLE, August 21st 2006 (583kb zip file)
Slides for GURU talk on Greeble Sonification, March 13th 2006 (1.8Mb .tar file)
Slides for Affective Effects AI seminar October 14th, 2005 (800kb .ppt file)
Slides for PEN Online Meeting April 20th, 2005 (2Mb .ppt file)
Audio feature extraction code for CSE253 lecture, Feb 29th 2008 (16kb zip file)
I built an online Harmonograph, a weird, wonderful way to visualize musical intervals