caffe-windows icon indicating copy to clipboard operation
caffe-windows copied to clipboard

Published 20 hours ago • verified publisher icon happynear

MTCNN model with memorydatalayer not efficient on ubuntu implemented caffe C++ code

Open Dicksonchin93 opened this issue 7 years ago • 1 comments

I tested your model which added in extra memorydatalayer and your bounding box regression layer. It seems that it takes longer time to perform the whole forward pass and outputs the final predicted bounding box location. The original model implementation in C++ is faster by 5 to 6 times. Using the densenet method does make it faster a bit though. Any thoughts on why is it slower?

Dicksonchin93 avatar Oct 27 '17 06:10 Dicksonchin93

I have tried without the memorydatalayer, and its still slow in initializing the net input with data and also slow in forward pass.

Dicksonchin93 avatar Oct 27 '17 08:10 Dicksonchin93