This first release contains the dataset and a trained mobilenetv2 model, that was trained on the three provided datasets and classifies scores with a precision of over 99%.
Additional, we publish the MusicScoreClassification dataset, that consists of 2000 images: 1000 images of music scores and 1000 images of other objects including text documents. The images were taken with a smartphone camera from various angles and different lighting conditions.