https://github.com/evhub/score-retrieval
Wrapper for music score retrieval dataset.
https://github.com/evhub/score-retrieval
Last synced: 10 months ago
JSON representation
Wrapper for music score retrieval dataset.
- Host: GitHub
- URL: https://github.com/evhub/score-retrieval
- Owner: evhub
- Created: 2018-10-13T22:16:49.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2019-05-04T23:52:58.000Z (over 6 years ago)
- Last Synced: 2025-01-30T01:44:56.652Z (12 months ago)
- Language: Python
- Size: 147 KB
- Stars: 2
- Watchers: 6
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Sheet ID
This repository contains the code used in "Sheet Music Identification Using Measure-Based CNN Features" by Hubinger, Khant, Kurashige, Amin, and Tsai.
## Installation
To install this repository, you will need to
1. `git clone https://github.com/evhub/score-retrieval.git`,
2. download the data to `/score-retrieval-data` or `./data`,
3. install [the CNN code](https://github.com/evhub/cnnimageretrieval-pytorch),
4. install [the measure segmentation code](https://github.com/aditya-khant/sheet-id-splitter), and
5. `make install`.
## Usage
First, you will need to ensure your data is properly set up. If you only have PDFs but not images, you will need to run `make pdfs-to-images`.
Then, to run the system, simply
```
python ./score_retrieval/run_all.py --alg
```
where `` is one of the following:
- `measure_segmentation` to run our best system using measure segmentation,
- `vgg_measure_segmentation` to run base vgg-gem with measure segmentation, and
- `tuned_measure_segmentation` to run measure segmentation utilizing a fine-tuned network.
To run `tuned_measure_segmentation`, you will need to have previously [run fine-tuning on the training data](https://github.com/evhub/cnnimageretrieval-pytorch).