🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

TODAY from 4PM: Challenges and Opportunities in Music Tech - open debate. We will discuss about the impact of music tech in the music sector with @robclouth, Marcelo Estrella, antònia folguera, Anna Xambó 🎗, Gonçal Calvo, Fabien Gouyon, Jordi Pons, Valerio Velardo upf.edu/web/mtg/home/-…

We did a paper for this #ISMIR2022, putting to use publicly available editorial metadata from my favorite music database, Discogs, to train audio models targeting music classification (400 styles) + contrastive learning by same track/release/artist/record label associations. 👇

At this #ISMIR2022, we presented a new benchmark dataset of relative arousal/valence annotations for validating audio models for music emotion recognition. 📄 Paper: hdl.handle.net/10230/54181 🔊 Dataset: mtg.github.io/musav-dataset/ w/ Xavier Lizarraga, Pablo Alonso Jiménez, Xavier Serra

Interested in language and music? 🎧💬Participate in Song Describer, a platform for collecting textual descriptions of music recordings upf.edu/web/mtg/home/-…

We added an option to download a lower-bitrate mono audio version of the MTG-Jamendo dataset (156 GB instead of 508 GB for over 55,000 full audio tracks labeled with almost 200 tags by genres, instruments, and moods/themes) and a fast download mirror. 💾🌐 mtg.github.io/mtg-jamendo-da…

“AI models for Sound and Music Applications” (AISMA), new Proof of Concept project at the MTG funded by Ministerio de Ciencia, Innovación y Universidades Agencia Estatal de Investigación #NextGenerationEU #PlanDeRecuperación: upf.edu/web/mtg/home/-…

We are looking for a senior researcher to join the MTG. Applications open until February 13th. Information here: upf.edu/web/mtg/home/-…

We're getting closer to having a CC-licensed music-text dataset collected through Song Describer, but we're still looking for contributions: song-describer.streamlit.app. From today, there will be new prizes for top contributors each week, so one more reason to get annotating ✏️🎹

Check out this article by Karma Peiró for a summary of our recent project at MTG-UPF surveying the opinions of over 100 music and tech professionals on the impact of AI and other emerging technologies. 👇 upf.edu/web/mtg/ai-and…

The MTG participates in #ICASSP2023 with six papers in which MTG researchers are involved: Benno Weck, Nazif Can Tamer, Pablo Alonso Jiménez, Dmitry Bogdanov, Pablo Zinemanas, Rajsuryan Singh, Christos Plachouras 🦋, Marius Miron, Lonce Wyse, Xavier Serra. upf.edu/web/mtg/home/-…

New paper at #ICASSP2023 led by Pablo Alonso Jiménez for his internship Utopia Music. It investigates using information about track sequences and co-occurrences in playlists for representation learning, building upon our previous work MTG-UPF that used simple editorial metadata.👇

We've updated Essentia with a new music representation model based on transformers and trained on music style metadata in Discogs. Useful for genre classification (400 styles) and downstream auto-tagging tasks. This work, led by Pablo Alonso Jiménez, will be presented at #ISMIR2023 👇

Replicate demo for tempo BPM estimation algorithms available in Essentia (degara, multifeature, percival, tempocnn): replicate.com/mtg/essentia-b…

We released the Song Describer Dataset of ~1.1K crowd-sourced music captions for 706 open-access audio tracks from MTG-Jamendo—work led by Ilaria Manco and Benno Weck, with Philip Tovstogan, Minz Won, and colleagues. Check our paper for the ML for Audio workshop at #NeurIPS2023. 👇

We've opened new PhD positions at MTG! Covering topics of sound and music representation learning, multi-modal music processing, machine listening, generative music, computational musicology, and music performance analysis. 👇🎓

I am happy to announce that we'll be presenting our music analysis software and AI models Essentia MTG and related activities of MTG-UPF at the Sónar 2024 Project Area, June 13 to 15. Sónar Festival Sónar+D 🔊⚡ w/ Pablo Alonso Jiménez, Jorge Marcos

New paper from our team for the #icassp2024 Workshop on Explainable AI for Speech and Audio.

The Song Describer Dataset is gaining some traction and is being used in the MIREX 2024 Music Description & Captioning task this year at #ISMIR2024.

New dataset of musical versions, built from public metadata on Discogs. We identified 1.9M versions across 348K music works, over 25% of which we mapped to official music uploads on YouTube. 💾 Dataset: mtg.github.io/discogs-vi-dat… 📄 Paper at #ISMIR2024: arxiv.org/abs/2410.17400

Find me on bsky.app/profile/dbogda…