Significant risks for the project we managed this week:
-
It’s almost certain that the originally proposed Bag-of-Words algorithm does not work well on the advanced midi matching task, and we are moving on to another approach that focuses more on the melody of the midi input, instead of the number of occurrences of each note.
- both Bag-of-Words with and without normalization, and the normalized version provides better results than the un-normalized version, yet the result is still not promising for the purpose of this project.
- The Bag-of-Words approach, even with normalization, tends to match more than one copies and the matched copies are ofter close to each other.
- We decide to move on to the Levenshtein distance, which should be a good choice to calculate the difference between two midi slices here. We are currently developing the algorithm described in this paper:
-
There’s a risk that since we decide to build a web application as interface for our users, and since none of us had experience with implementing that for mobile users (specifically the part for uploading files since pictures are taken with phones and not computers), there’s a chance that it will not be compatible with mobile devices or that it will take a longer time than expected to build the web application
Changes made to the existing design of the system:
-
We are now unsure if our originally proposed computer vision approach to recognize notes/signs on the sheet music will work. We found two reliable resources here and here. We will try these out but if the implementation takes more time than necessary, we will move on to manual labeling to ensure that we are on schedule.
An updated schedule if changes have occurred:
-
N/A
0 Comments