Recent advancements in music AI are focusing on enhancing interpretability, data generation, and genre-specific applications. New datasets like ConceptCaps provide structured, labeled examples for better model training, while innovative approaches in automatic drum transcription are leveraging semi-supervised methods to create high-quality datasets from unlabeled audio, addressing the scarcity of paired audio-MIDI data. Additionally, the introduction of models like D3PIA for piano accompaniment generation demonstrates improved coherence through localized attention mechanisms. In the realm of electronic music, EDMFormer employs self-supervised learning tailored to genre-specific characteristics, improving segmentation accuracy for EDM tracks. Meanwhile, efforts in music plagiarism detection are gaining traction, with new frameworks and datasets aimed at clarifying the task and enhancing practical applications. These developments collectively aim to solve commercial challenges in music production, copyright enforcement, and automated music analysis, reflecting a shift towards more specialized and efficient AI solutions in the music industry.
Concept-based interpretability methods like TCAV require clean, well-separated positive and negative examples for each concept. Existing music datasets lack this structure: tags are sparse, noisy, or ...
Deep learning models define the state-of-the-art in Automatic Drum Transcription (ADT), yet their performance is contingent upon large-scale, paired audio-MIDI datasets, which are scarce. Existing wor...
Music structure segmentation is a key task in audio analysis, but existing models perform poorly on Electronic Dance Music (EDM). This problem exists because most approaches rely on lyrical or harmoni...
Generating piano accompaniments in the symbolic music domain is a challenging task that requires producing a complete piece of piano music from given melody and chord constraints, such as those provid...
In recent years, foundation models have become very popular due to their exceptional performance, mainly in natural language (NLP) tasks where they were first introduced. These models usually consist ...
Recently, the problem of music plagiarism has emerged as an even more pressing social issue. As music information retrieval research advances, there is a growing effort to address issues related to mu...
Combining multiple audio features can improve the performance of music tagging, but common deep learning-based feature fusion methods often lack interpretability. To address this problem, we propose a...
Motifs often recur in musical works in altered forms, preserving aspects of their identity while undergoing local variation. This paper investigates how such motivic transformations occur within their...