It is worth noting that AcousticBrainz was based on an open-source audio analysis library Essentia [1] that has been gradually improving since the launch of the AcousticBrainz project. It now has better ML models, with higher accuracy and generalization, than those used in AcousticBrainz in the pre-deep learning age of handcrafted signal processing features and SVMs.
See [2] for the list of models currently available. It includes updated ML models for the classifiers used in AcousticBrainz and many new models (e.g., a 400-style music classifier, models for arousal-valence emotion recognition, and feature embeddings that can be used for audio similarity or training new classifiers).
As a part of an ongoing research collaboration between the Centre for Digital Music of the Queen Mary University of London and the Music Technology Group of Universitat Pompeu Fabra Barcelona, we are happy to announce the launch of Song Describer, a crowdsourcing initiative to collect the first open research dataset of music captions.
Song Describer is an open-source annotation platform featuring CC-licensed music that anyone can annotate with natural language descriptions. Through this platform, we aim to create the first research dataset of music-caption pairs, which will give us insights into how people describe music and support the development of audio-text ML models.
There is an increasing need for music datasets that go beyond categorical labels and we hope that our data collection efforts will promote more research on the relationship between natural language and music audio, and its many potential applications. If you're interested in this field or are curious or willing to contribute, you can head to Song Describer to read more about it and start annotating right away!
hledger is another open-source CLI tool for accounting that works amazingly well for me. You can add transactions editing in plain text too and there is a basic web UI on top of it.
Hi, I'm another researcher at the Music Technology Group, which runs Freesound. Definitely, this one of our long-time challenges.
Facilitating search and exploration is one of our main research directions related to Freesound. We do audio ML research, in particular, audio-based auto-tagging (see the already mentioned Freesound Datasets https://annotator.freesound.org/fsd/ we've built for this purpose) but also similarity and clustering based on audio analysis and tags.
So far, Freesound API provides similarity search and audio analysis features.
See [2] for the list of models currently available. It includes updated ML models for the classifiers used in AcousticBrainz and many new models (e.g., a 400-style music classifier, models for arousal-valence emotion recognition, and feature embeddings that can be used for audio similarity or training new classifiers).
[1] https://essentia.upf.edu/ [2] https://essentia.upf.edu/models.html