README.md

---
license: apache-2.0
language:
- en
library_name: transformers
pipeline_tag: image-to-text
---

# Optical Music Recognition Transformer

<!-- Provide a quick summary of what the model is/does. [Optional] -->
Image-To-Text model for optical music recognition. 
The model is trained to predict simple notes in the [LilyPond](https://en.wikipedia.org/wiki/LilyPond) format from a given image. 
Training data consists of artificial, handwritten and white board images. 
The model itself is based on [Donut](https://huggingface.co/docs/transformers/model_doc/donut).

## Demo

![White Board Sample](sample1.png)

Prediction: `c'2 a''8 c''8 r4 c'1 e'8 c'8 c'8 a''8 f'4 a'8 c'8`


![White Board Sample](sample2.png)

Prediction: `d'8 g'8 c''8 a'8 d'2 c'8 f''8 d'4 c''4 e'8 r8 g'8 b'8 e'8 g'8 d'2`


Repo: https://github.com/UHHRobotics22-23/robot_project/tree/main/marimbabot_vision

Note: For historical reasons, images need to be rotated by 90 deg to the left to get the best performance. This is also the case for the "Hosted inference API".