Skip to content

/bird-recognition

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
2nd Place Model for Kaggle Competition
  1. Python 100%
Python
branch: master
bird-recognition/

Initial commit

latest commit efc08a5891
Matt Wescott authored
Failed to load latest commit information.
FoldBuilder.py Initial commit
README.md Initial commit
SETTINGS.json Initial commit
classifier.py Initial commit
config.py Initial commit
create_submission.py Initial commit
data.py Initial commit
mfccs.py Initial commit
requirements.txt Initial commit
sanity_check.py Initial commit
submission_utils.py Initial commit
training_utils.py Initial commit
transform.py Initial commit
wav_utils.py Initial commit

README.md

Bird Species Classification

Hardware / OS

  • M3.2XLarge instance on Amazon EC2
    • 8 virtual cpu
    • 26 gb ram
    • Intel Xeon E5-2670
  • Ubuntu 12.04, 64 bit, Server Edition

Dependencies

  • numpy/scipy
  • pandas
  • scikit-learn

Installation Instructions (from fresh EC2 instance)

  1. sudo apt-get update
  2. sudo apt-get install git
  3. sudo apt-get install python-pip python-dev build-essential libblas-dev liblapack-dev gfortran
  4. sudo pip install virtualenv
  5. git clone git@github.com:mattwescott/bird-recognition.git
  6. cd bird-recognition
  7. Get the data
  8. Edit paths in SETTINGS.json to point to the data
  9. virtualenv venv
  10. source venv/bin/activate
  11. pip install -r requirements.txt

Generate Kaggle submission

  • python create_submission.py (~4 hours runtime)

Predict on new test set

  • Not implemented
Something went wrong with that request. Please try again.