BlueAnthony

Ur code only add output with input when the stride is 1. How about that the stride is 2?

As title, ur code only do the shortcut when stride is 1. And if the stride is 2, it doesnt add anything.

Resnet Shortcut Condition

1

想请问一下这行 if stride != 1 or in_planes != self.expansion*planes: 这两个条件看起来不可能单独存在 两个条件一定会同时符合，所以判断其中一个似乎就够了？ 也就是说，不会有下列几种可能： 1. stride != 1 但 in_planes == self.expansion*planes 2. stride == 1 但 in_planes != self.expansion*planes 还是我哪里理解错了？

Does this mean that caffe-ssd needs to be compiled with caffe and tensorRT ssd only adopts the layers from tensorRT without caffe?