Skip to content

ThomasRusbjerg/sw10

Repository files navigation

DETR for Combined Object Detection and Notation Assembly in Optical Music Recognition

This project was about using the DETR architecture to identify and classify musical objects and their relations in images of handwritten sheet music (in western music notation). The dataset used can be found here: MUSCIMA++. Our main contribution was extending the DETR architecture to be able to classify the relations between objects such that end-to-end object detection and notation assembly can be handled holistically.

Repo content

This repository contains the code used to train and evaluate our model. The training was partly done on a AAU provided Nvidia DGX System and on Google Cloud using the AI Platform. The thesis report is located in the repository as well mi1014f21_thesis.pdf.

About

Master Project - AAU - Software - Machine Intelligence Specialisation

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages