search

ShiqiYu / libfacedetection.train

The training program for libfacedetection for face detection and 5-landmark detection.
Apache License 2.0
774 stars 209 forks source link

请问余老师如何理解网络结构中ConvDPUnit #28

Open XuhongLuo opened 3 years ago

XuhongLuo commented 3 years ago

请问余老师ConvDPUnit中的卷积部分如下代码： image

为什么没有使用pointwise的操作进行降维呢？这样操作对速度和精度有什么作用吗？谢谢

ShiqiYu commented 3 years ago

降了的话，还要再升。这样简洁。

On Jul 14, 2021, at 14:59, XuhongLuo @. @.>> wrote:

请问余老师ConvDPUnit中的卷积部分如下代码： https://user-images.githubusercontent.com/35533084/125576661-22b8e6c7-bf4c-4d6b-8319-fccd9df38aa1.png 为什么没有使用pointwise的操作进行降维呢？这样操作对速度和精度有什么作用吗？谢谢

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ShiqiYu/libfacedetection.train/issues/28, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABWR4HM4EWVGDBFOM7LUI43TXUYTNANCNFSM5AK2KZOA.

XuhongLuo commented 3 years ago

@ShiqiYu 嗯嗯，这样的话网络速度会有提升，那余老师人脸检测效果相对降维的话，差异明显吗？