The article presents BEAST (BEA t tracking Streaming Transformer), a system for online beat and downbeat tracking. The system, which is based on the streaming Transformer, is designed to handle online scenarios and uses contextual block processing in the Transformer encoder. The system also uses relative positional encoding in the attention layer of the streaming Transformer encoder to capture relative timing position in music. In tests, BEAST showed a significant improvement over the current state-of-the-art online beat and downbeat tracking model.
Publication date: 29 Dec 2023
Project Page: Unavailable
Paper: https://arxiv.org/pdf/2312.17156
