What is the speed of Mobilenet-v2 running on caffe?
AlexofNTHU opened this issue · 1 comments
AlexofNTHU commented
Using my TitanX(Maxwell), I got 20ms for a 224*224 image inference with batch=1.
How fast could you achieve on caffe?
Thanks,
Alex
austingg commented
@AlexofNTHU sorry， I have not used caffe for a long time. If you use cuda and cudnn, the inference speed will be much faster than 20 ms. Maybe you need some warmup and get average inference time.