CLaMP 3 revolutionizes music retrieval across languages and modalities using cutting-edge contrastive learning techniques.
更新时间:2025-02-20 21:40:40
CLaMP 3 is a pioneering framework in music information retrieval (MIR), addressing the challenges of cross-modal and cross-lingual music search. By utilizing contrastive learning, CLaMP 3 can align various music representations—including sheet music, audio, and performance signals—across different languages and modalities. This allows users to retrieve music with high accuracy using textual descriptions in multiple languages. Through its innovative approach, CLaMP 3 enables music retrieval even when traditional methods fail due to the unavailability of aligned data.
To use CLaMP 3, users can input queries in various languages to retrieve corresponding music across different formats such as sheet music, audio, or performance signals. The framework employs a multilingual text encoder to process queries and match them to music representations. Additionally, users can also perform cross-modal music retrieval, such as finding audio or sheet music that corresponds to a specific image or video scene, enhancing the flexibility of music search.
CLaMP 3目前提供免费在线演示,用户可以通过其Hugging Face空间直接体验。关于API和企业级服务的具体价格,需联系CLaMP 3团队获取更多定价信息。
CLaMP 3由多所知名学术机构联合研发,包括中央音乐学院、香港科技大学、纽约大学上海校区、穆罕默德·本·扎耶德人工智能大学、韩国高等科学技术学院及清华大学。
CLaMP 3的Email地址:[[email protected]](mailto:[email protected])
CLaMP 3的社交媒体: - Twitter: @CLaMP3_MIR - Instagram: @clamp3_music