Jiameng’s Status Update for 02/15/2020

Accomplishments this week

• Tried out methods for translating a music score into midi format data
  • Initially tried to use the software ScanScore and though it is quite useful, it has a couple of problems
    • Advantage: The recognition of the software is quite accurate though need to proofread through the output to make sure the midi matches the music score, also, the process of recognition is very quick. You can hear the music score being played so that it would be easier to see if the music score is transformed correctly.
    • Disadvantage: To use it, need to find high resolution music scores and transform pdf formats into tiff format in order for the software to recognize it. There is also a chance of failure for reading the input music score. It is not very convenient to output the midi data as it only outputs MusicXML and need to find other websites or software in order to transform MusicXML into midi.
  • Decided to download both music scores and its corresponding midi data from the website musescore and store those into database. Reason for choosing this is for the purpose of demonstrating our project, the number of music scores in database is not the most important for us and currently, it’s fine to just download scores and midi files. Also, in the future, if we want to expand the database, we can write Python programs to fetch that url and automate the process of downloading the two types of files.
  • Currently, the local database of our project consists of 20 music scores and 20 corresponding midi files, all files contains at least 2 pages so that we can use these to test our page flipping functionality
• Figured out how to use the cv2 library of python to adjust the picture of music scores taken by phones
  • The music score picture taken by phone could be twisted and not in a rectangular shape which makes matching with database harder, so need to use camera calibration methods to match the four corners of the picture taken with the pictures stored in database (which are pdfs in regular formats)
  • Currently, is able to use software on phone like Office Lens to take pictures that look like photocopy ones and the picture should not be too different from the music score, later on might use machine learning algorithms to raise the accuracy and success rate of matching the music score with those in database

Progress for schedule

• On schedule

Deliverables I hope to accomplish next week

• Finish camera calibration to match a music score in the database
• Start matching an input music staff with music staff in the database