caffe mobilenet-ssd测试demo.py报错Ch

原因：

original Caffe 

It looks like that Batchnorm layer and adjacent Scale layer are 

integrated in single Batchnorm layer in NVIDIA caffe. However original 

BVLC/caffe doesn’t integrate them. That means any networks and 

pretrained models using batchnorm and scale layers with original Caffe 

cannot use in NVIDIA caffe and DIGITS.

即：

包含：Batchnorm层和adjacent Scale层分开的； 

NVIDIA Caffe/DIGITS包含：Batchnorm层和adjacent Scale层合并在一个BN层中； 

所以在预测过程中不能使用NVIDIA Caffe，也就是GPU。

---

解决方法：修改deploy.prototxt

左边为修改前，右边为修改后