BriefGPT.xyz
Aug, 2023
动态nsNet2: 高效的深度噪声抑制与提前退出
Dynamic nsNet2: Efficient Deep Noise Suppression with Early Exiting
HTML
PDF
Riccardo Miccini, Alaa Zniber, Clément Laroche, Tobias Piechowiak, Martin Schoeberl...
TL;DR
我们提出基于nsNet2的早期退出模型，通过在不同阶段停止计算，提供多个精度级别和资源节省，并通过分流信息流来适应注入的动态，展示了性能和计算复杂性之间的权衡。
Abstract
Although
deep learning
has made strides in the field of
deep noise suppression
, leveraging deep architectures on
resource-constrained devices
→