Mask RCNN裁剪后移植到Android端，报错解决方案

模型是在tensorflow r1.12中训练的，所以使用tensorflow r1.12的源码编译libtensorflow_inference.so来配套移植到Android端，结果报了以下两个错误：

1. Op type not registered 'NonMaxSuppressionV3' in binary running on localhost. Make sure the Op and Kernel are registered in the binary running in this process.

参考：https://www.jianshu.com/p/056ba3cbf02a

2. No OpKernel was registered to support Op 'DenseToDenseSetOperation' with these attrs.  Registered devices: [CPU], Registered kernels:

1) 在tensorflow/contrib/makefile/tf_op_files.txt中添加set_kernels.cc

2) 在tensorflow/core/kernels/BUILD中找到：

filegroup(
    name = "android_extended_ops_group2",
    srcs = [

    ]

在srcs里面加上set_kernels.cc

3) 重新编译libtensorflow_inference.so