Rookie_Chris

The dimension of mbox_priorbox is wrong, and you can check it again. @paghdv

There is the vgg16-pruning prototxt in https://github.com/yihui-he/channel-pruning, and what you need to do is to combine it with SSD.

It mainly reduces some of the channels and does not add some layers. If your precision is not good enough, you can choose to add some bn and scale layers....

Okay, I see what you mean. I will organize a tutorial.

This tutorial is not very well written. I perform channel pruning on ssd by referring to the SSD+VGG format .@tianfanzhu

I had a similar question that invalid caffe_.mex64. Do you compiled ShaoqingRen's caffe in cpu model?

can u provide your caffe_mex64 of the cpu version?Thank u very mech. My gmail is [email protected].

How much faster the speed of fcos with tensorrt than original version? @jinfagang