Accomplishments this week
- Cleaned up CSS for the website
- Centering buttons and texts
- Changed UI for each page
- Added a button so that when the button is clicked, an explanation of what the page is used for will show up, and another click will hide the explanations
- A button on the display page that when a user click it, he can manually flip the page to the left or to the right
- Completed testing for score matching program, which is the program on web app that when user inputs a picture of score, recognizes its title and then uses that title to match the ones in database
- For title recognition, since I don’t have a printer, I’m mostly testing with the original pdfs with some modifications.
- Originally, accuracy of title recognition is about 90% since the title and the file name stored in database does not exactly match, and there are parts in different scores that look the same, for example “easy version” exists in multiple scores, so, to increase accuracy, I changed the filenames in database to match those on the original scores and changed the name of the midi files accordingly
- Currently, with modified scores (such as blurring a little bit), accuracy of recognition reaches 100%, with pictures taken by photo, the the text printed does not exactly match the title but with a small sample of scores in database, accuracy of matching still reaches 100%
- Completed testing for score recognition
- For score recognition, the accuracy currectly is 94.86% for normal scores and with simpler ones, can reach 100%, for the ones that we choose to include in our database, the accuracy is around 97% so we still need to make some modifications to the midi file, so that the number of notes before a flip point is correct
- Finished research and decided instance type for aws
- aws provides us with different instance types, and in order to push our service to the instances, it is important that we choose the right instance to deploy our service, so we need to trade off between performance and cost
- instances includes A1, T3, T3a … which are suited for different tasks. According to research, I choose type T3 because it provides ample workload support and when a workload is operating below baseline threshold, it could also save credit, so it is useful
- to ensure performance and storage of data, need to choose between t3.large and t3.medium and we still need to do some testing next week to figure out which one we want to use
Progress for schedule
- On schedule
Deliverables I hope to accomplish next week
- Push webapp to one aws instance and MIDI tracking program to another instance
- Start working on the connection between the two instances
0 Comments