Researchers Develop Comprehensive Dataset for Portuguese Folk Instruments

Strategic Insights -

A team of researchers from the University of Salamanca in Spain has created a comprehensive dataset of visual data on traditional Portuguese musical instruments. The dataset, aimed at supporting computer vision and artificial intelligence applications, contains images of various folk instruments used in Portugal's traditional music. By making this dataset publicly accessible, the researchers hope to facilitate ongoing research in computer vision, cultural preservation, and cataloguing of cultural heritage.

Key Takeaways:

  • The dataset contains visual data of 15 traditional Portuguese musical instruments extracted from videos available on the "webpage", a cultural and musical documentation platform.
  • Each image highlights the distinctive visual characteristics of the instruments as they are being played, and the dataset has been standardised to a uniform resolution without applying additional filters.
  • The dataset includes a variety of traditional musical instruments used in Portuguese folk music, offering a structured resource for training and testing computer vision models aimed at instrument recognition and classification.
  • The dataset has been labelled using the Roboflow tool, and the images are organised to support the creation of machine learning models and artificial intelligence applications.
  • The research aims to facilitate ongoing research in computer vision, cultural preservation, and cataloguing of cultural heritage.
  • The dataset is publicly accessible, providing a detailed visual tool for comparative study and classification of culturally significant instruments.
  • The researchers include Hector Sanchez San Blas, Luis Augusto Silva Zendron, Andre Filipe Sales Mendes, Diego M. Jimenez-Bravo, and Alvaro Lozano Murciego as authors of the research.

Statistics:

  • The dataset contains 15 traditional Portuguese musical instruments.
  • The dataset has been standardised to a uniform resolution of 1280x720 pixels.
  • The dataset includes 500 images of each instrument, with a total of 7500 images.
  • The dataset has been labelled using the Roboflow tool.
  • The research aims to support train and test computer vision models for instrument recognition and classification.

Sources:

  • Data In Brief, 2025;61:111739.
  • NewsRx. University of Salamanca Reports Findings in Artificial Intelligence (Comprehensive dataset of Portuguese folk instruments for computer vision and heritage research). Computer Weekly News. July 16, 2025; p 833.