1 post

  • avatar
    0 sounds
    2 posts

    Freesound Datasets is a platform that we’ve been developing to foster the re-use of Freesound content in research contexts. Sounds familiar? This is because we already introduced you to Freesound Datasets in this blog post. The goal of Freesound Datasets is to create datasets, that is to say, collections of sounds annotated with predefined labels. Well-curated datasets are a key factor in audio research and we intend to do research on applications that will eventually make Freesound better (yay!).

    Our main objective with Freesound Datasets is to create the FSD, a very large dataset for doing research in sound recognition. But since this is a long term thing, we thought that in the meantime we could share a small version of it. So far we've compiled a first small version, and we've organized a competition in Kaggle with it (see Freesound General-Purpose Audio Tagging Challenge). In this competition, participants are challenged to build artificial intelligence algorithms able to recognize 41 categories of everyday (Free)sounds. Shortly after the competition is over, at the end of the summer, we’ll make this dataset version available for download, including lists of Freesound clips annotated by awesome contributors and curated by us at the MTG. Among the categories included in this first release, you will find from musical instruments to human sounds, as well as a few domestic sound categories and bit of miscelanea.

    The (Free)sound collections in Freesound Datasets are work-in-progress and their development depends heavily on the good will of users that contribute with annotations. We invite you to join the good cause by visiting the platform and contributing with some sound annotations. While you can do that anytime by simply visiting this page, we will also organize short-term campaigns in Freesound to gather annotations with a specific goal. Say, a couple of weeks focused on accomplishing an annotation target. Oh, and we plan to reward top annotators with Freesound t-shirts!

    This new forum section, "Freesound Datasets", has been created with the goal of getting feedback and having discussions with those Freesound users interested in the platform, and to provide you up to date information about the project. Hence, feel free to share your thoughts and feedback about Freesound Datasets (and/or any particular sound collection in it) here. Also don’t forget to join our mailing list if you want to get updates delivered to your inbox smile

    1 post