AutoOsu is an end-to-end rhythm game chart generation model. Given an audio track and BPM/offset values, it generates a 'beatmap' file that can be played on the game 'osu! mania'. While it's still very much a work in progress, I'm excited to showcase it at #ISMIR2023 LBD!
1/ 📢 Exciting news from our latest #ISMIR2023 paper, “On the Effectiveness of Speech Self-supervised Learning for Music”🎉 We've explored the application of Self-Supervised Learning (SSL) models, including data2vec and HuBERT, originally designed for speech to music.
1/ 📢 Introducing our #ISMIR2023 paper, LyricWhiz! 🎉 LyricWhiz is the first system to perform zero-shot, multilingual, long-form lyrics transcription by integrating Whisper and GPT4 from OpenAI.
Paper: arxiv.org/abs/2306.17103
Code&Dataset: github.com/zhuole1025/Lyr…
/
Excited to announce that our state-of-the-art piano transcription model, the 'hFT-Transformer'🎹, will be presented at the #ISMIR2023 international conference in Italy in just two weeks.😉🎻🎷🎵
\
🔹paper 🪗: arxiv.org/abs/2307.04305
🔹code🎸: github.com/sony/hFT-Trans…
Honored to…
![/
Excited to announce that our state-of-the-art piano transcription model, the 'hFT-Transformer'🎹, will be presented at the #ISMIR2023 international conference in Italy in just two weeks.😉🎻🎷🎵
\
🔹paper 🪗: arxiv.org/abs/2307.04305
🔹code🎸: github.com/sony/hFT-Trans…
Honored to…](https://pbs.twimg.com/media/F9Mnc8mb0AA2etx.jpg "Taketo Akama @Tokyo🍣🗼🗻🏯🇯🇵 (@tuktoe) on Twitter photo 2023-10-24 10:11:32 /
Excited to announce that our state-of-the-art piano transcription model, the 'hFT-Transformer'🎹, will be presented at the #ISMIR2023 international conference in Italy in just two weeks.😉🎻🎷🎵
\
🔹paper 🪗: arxiv.org/abs/2307.04305
🔹code🎸: github.com/sony/hFT-Trans…
Honored to…")
#ISMIR2023 hot take #2: much love for the efforts to facilitate further research! Lots of pip installable software, pre-trained models and new datasets. Also some great efforts with training smaller models and using self-supervised learning 💯
Second place in AIcrowd's challenge presented at #ISMIR2023 🥳Thank you for organising Yuki Mitsufuji moises.ai Sony FaRo Alexandre Défossez
#ISMIR2023 hot take #3: AI trained on music of different cultures has a better understanding of western music than AI trained only on that domain, and vice versa. I’d say it’s very similar for humans as well. One of many fascinating musical insights from ISMIR Conference 🔥
My supervisors were taking photos of me presenting at #ISMIR2023 . One in Milan Juhan Nam one from Tokyo Taketo Akama @Tokyo🍣🗼🗻🏯🇯🇵 🤩
Tomorrow, I will present at #ISMIR2023 : Predicting performance difficulty from piano sheet music images. Our method addresses a more realistic scenario than previous ones: sheet music images 🎶📝 #EdTech #MIR Dasaem Jeong MTG-UPF Jose Javier Valero Mas.
Paper: arxiv.org/abs/2309.16287
⅙
