Multi-Task Multi-Frame Visual Piano Transcription

  • 类型:arxiv
  • 标识:2608.03419
  • 链接:https://arxiv.org/abs/2608.03419
  • 主分类:multimodal
  • 形态:method
  • 被引:0
  • 被引来源:Semantic Scholar
  • S2被引:0
  • 影响力被引:0
  • TLDR:This work presents V2N (Video to Notes), the first complete VPT system: a shared temporal backbone feeds task-specific heads for onset, offset, key hold, and velocity, jointly trained with per-frame supervision rather than only at the window center.
  • 待LLM分类:否
  • 标题中文:多任务多帧视觉钢琴转写
  • TLDR中文:提出 V2N (Video to Notes),首个完整的 VPT 系统:共享时序 backbone 驱动 onset、offset、key hold、velocity 等任务专属 head,采用逐帧监督而非仅在窗口中心进行联合训练。
  • 来源文件
  • /inbox/tom/_candidates/2026-08-05-agent-rag-longcontext-candidates.json
  • [S2 enrich]