caffemodel转换为tensorflow的pb格式

近期要因为学习的需要要从ｃａｆｆｅ框架转为ｔｅｎｓｏｒｆｌｏｗ的学习，但是通过在网上搜索后发现，网上关于训练好的模型却几乎找不到，因此心里在想，有没有相关的代码将ｃａｆｆｅ训练好的模型转换为ｔｅｎｓｏｒｆｌｏｗ的模型呢？说干就干，在网上搜索了一下，还真的找到了一个相关的开源代码，可以实现这个功能.其在ｇｉｔｈｕｂ上为开源工程，链接为：https://github.com/ethereon/caffe-tensorflow
于是从网页上ｇｉｔ下来了这个项目源码，按照工程中提示的信息一步一步的进行操作，可是在刚开始的时候就给了当头一棒，泼了一盆冷水，当按照教程中提示的指令输入
./convert.py examples/mnist/lenet.prototxt --code-output-path=mynet.py
最不想开到的情况还是出现了，程序报错．

Traceback (most recent call last):
  File "./convert.py", line 60, in < module> main()
  File "./convert.py", line 56, in main args.phase)
　File "./convert.py", line 27, in convert transformer = TensorFlowTransformer(def_path, caffemodel_path, phase=phase)
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/tensorflow/transformer.py", line 221, in __init__ self.load(def_path, data_path, phase)
  File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/tensorflow/transformer.py", line 227, in load graph = GraphBuilder(def_path, phase).build()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/graph.py", line 140, in __init__ self.load()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/graph.py", line 144, in load self.params = get_caffe_resolver().NetParameter()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/resolver.py", line 32, in get_caffe_resolver SHARED_CAFFE_RESOLVER = CaffeResolver()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/resolver.py", line 7, in __init__ self.import_caffe()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/resolver.py", line 17, in import_caffe from . import caffepb
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/caffepb.py", line 28, in <module>type=None),
　File "/usr/local/lib/python2.7/dist-packages/google/protobuf/descriptor.py", line 654, in __new__ message.Message._CheckCalledFromGeneratedFile()
TypeError: Descriptors should not be created directly, but only retrieved from their parent

在网上搜索了一下，并没有发现合适的方法，最终还是选择了ｇｏｏｇｌｅ大神来帮忙解决这个问题，网上给出了这样一个解决方法，就时将kaffe/caffe文件夹中的caffepb．py改为caffe_pb2.py,原本以为这样感觉不太合理啊，只是改一下名字就能解决这个问题？但是也没别的什么办法，就试一下看看具体会报什么错误吧．
于是又重新输入了一遍指令，程序果然报错了，提示：

Traceback (most recent call last):
 File "./convert.py", line 60, in <module>　main()
 File "./convert.py", line 56, in main　args.phase
 File "./convert.py", line 27, in convert
    transformer = TensorFlowTransformer(def_path, caffemodel_path, phase=phase)
 File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/tensorflow/transformer.py", line 221, in __init__　self.load(def_path, data_path, phase)
 File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/tensorflow/transformer.py", line 227, in load　graph = GraphBuilder(def_path, phase).build()
 File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/graph.py", line 140, in __init__　self.load()
 File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/graph.py", line 144, in load
    self.params = get_caffe_resolver().NetParameter()
 File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/resolver.py", line 32, in get_caffe_resolver　SHARED_CAFFE_RESOLVER = CaffeResolver()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/resolver.py", line 7, in __init__　self.import_caffe()
　File "/home/li/liyahui/caffe/caffe-tensorflow/kaffe/caffe/resolver.py", line 17, in import_caffe
    from . import caffepb
ImportError: cannot import name caffepb

按照程序提示的错误信息，找到文件reolver.py这个文件，并定位到from . import caffepb的位置，由于文件名修改了，那必定找不到这个文件，更不用谈导入相关程序代码了，所以程序必定报错．在明白了这个之后将from . import caffepb改为

 # Fall back to the protobuf implementation
        from . import caffe_pb2
        self.caffepb = caffe_pb2
        show_fallback_warning()

再运行命令，发现成功了．程序运行出了结果，生成了mynet.py文件，其内容为:

from kaffe.tensorflow import Network

class LeNet(Network):
def setup(self):
   (self.feed('data')
    .conv(5, 5, 20, 1, 1, padding='VALID', relu=False, name='conv1')
    .max_pool(2, 2, 2, 2, name='pool1')
    .conv(5, 5, 50, 1, 1, padding='VALID', relu=False, name='conv2')
    .max_pool(2, 2, 2, 2, name='pool2')
    .fc(500, name='ip1')
    .fc(10, relu=False, name='ip2')
    .softmax(name='prob'))

caffemodel转换为tensorflow的pb格式

猜你喜欢