The cudaMemcpy is time consuming

ChenYingpeng / caffe-yolov3

A real-time object detection framework of Yolov3/v4 based on caffe

473 stars 231 forks source link

The cudaMemcpy is time consuming #34

Closed wywlzyzz closed 5 years ago

wywlzyzz commented 5 years ago

I added a for loop in the main function to detect images, but the time consumption of cudaMemcpy in the cuda_make_array function located in cuda.cpp increased sharply, about 200 times, after the first detection.

Is there any method to solve this problem? THX.

BoyangYu96 commented 5 years ago

excuse me, how much time do you cost to detect average picture? @wywlzyzz