We set up a Hg repository today for the project at bitbucket.org. Michael and I looked around the web for Python wrappers for OpenCV, but the ones that we found were either deprecated, poorly documented, or both. I installed a Java wrapper for OpenCV called JavaCV on my laptop. It appears to work properly and that may be what we use.
We also looked a little bit into the algorithms. For music OCR, one of the most important steps is staff recognition and removal. We read about a few common algorithms used to accomplish this task, but we are still undecided on which one to use. Another important step for music OCR is symbol segmentation. This comes after staff recognition and removal; after reading a paper, we have picked the k-NN technique for its simplicity and effectiveness.
No comments:
Post a Comment