Skip to content
Hot eventLive

脉冲间隔变化对蝙蝠叫声深度学习分类影响研究

1 reports1 sources16 hr ago updated

Get the story

AI overview

2026年10月5日,arXiv Machine Learning Theory发布一手研究,比较脉冲间隔(IPI)变化对蝙蝠叫声物种分类的影响。研究在欧洲蝙蝠录音上构建自然IPI与50 ms归一化IPI两组数据。结果显示,PaSST在两种条件下准确率分别为71±2.3%与70±6.3%,EfficientNet-B0从47±4.7%升至57±3.9%,PaSST在两种条件下均占优;BatDetect2与BAT的差异不大。

Generated from reports · updated 16 hr ago

Timeline

Follow the coverage from different angles.

Oct 5, 2026
  1. arXiv · Machine Learning Theory
    脉冲间隔变化对深度学习蝙蝠叫声分类的影响

    研究比较了脉冲间隔(IPI)变化对蝙蝠叫声物种分类的影响,在欧洲蝙蝠录音上构建自然 IPI 与 50 ms 归一化 IPI 两组数据。PaSST 在两种条件下准确率分别为 71±2.3% 与 70±6.3%,EfficientNet-B0 从 47±4.7% 升至 57±3.9%,PaSST 均占优;BatDetect2 与 BAT 差异不大。

Heat trend

Current heat 7·Comparable peak 10(Oct 5)·Comparable change over 24 hours –

02.557.510Oct5Oct5Oct5Oct6

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.