Caffe - GoogLeNet 忽略源层答案

【问题标题】：Caffe - GoogLeNet Ignoring source layerCaffe - GoogLeNet 忽略源层
【发布时间】：2018-02-27 00:35:27
【问题描述】：

我在docker 容器内运行Caffe 框架（CPU），代码如下：

model_path = 'models/bvlc_googlenet/' # substitute your path here
net_fn   = model_path + 'deploy.prototxt'
param_fn = model_path + 'bvlc_googlenet.caffemodel'

# Patching model to be able to compute gradients.
# Note that you can also manually add "force_backward: true" line to "deploy.prototxt".
model = caffe.io.caffe_pb2.NetParameter()
text_format.Merge(open(net_fn).read(), model)
model.force_backward = True
open('models/bvlc_googlenet/tmp.prototxt', 'w').write(str(model))


net = caffe.Classifier('tmp.prototxt', param_fn,
                       mean = np.float32([104.0, 116.0, 122.0]), # ImageNet mean, training set dependent
                       channel_swap = (2,1,0)) # the reference model has channels in BGR order instead of RGB

但是，在忽略所有模型层之后，我得到以下回溯：

libdc1394 error: Failed to initialize libdc1394
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0227 00:08:49.033560   122 net.cpp:49] Initializing net from parameters: 
force_backward: true
state {
  phase: TEST
}
I0227 00:08:49.033826   122 net.cpp:274] Network initialization done.
I0227 00:08:49.316442   122 upgrade_proto.cpp:52] Attempting to upgrade input file specified using deprecated V1LayerParameter: models/bvlc_googlenet/bvlc_googlenet.caffemodel
I0227 00:08:49.370679   122 upgrade_proto.cpp:60] Successfully upgraded file specified using deprecated V1LayerParameter
I0227 00:08:49.374241   122 net.cpp:752] Ignoring source layer data
I0227 00:08:49.374426   122 net.cpp:752] Ignoring source layer label_data_1_split
I0227 00:08:49.374480   122 net.cpp:752] Ignoring source layer conv1/7x7_s2
I0227 00:08:49.374514   122 net.cpp:752] Ignoring source layer conv1/relu_7x7
I0227 00:08:49.374563   122 net.cpp:752] Ignoring source layer pool1/3x3_s2
I0227 00:08:49.374616   122 net.cpp:752] Ignoring source layer pool1/norm1
I0227 00:08:49.374683   122 net.cpp:752] Ignoring source layer conv2/3x3_reduce
(for all layers)
Traceback (most recent call last):
  File "dream.py", line 40, in <module>
    channel_swap = (2,1,0)) # the reference model has channels in BGR order instead of RGB
  File "/root/caffe/python/caffe/classifier.py", line 29, in __init__
    in_ = self.inputs[0]
IndexError: list index out of range

谁能指点我正确的方向？

【问题讨论】：

标签： python caffe

【解决方案1】：

这行得通：

model_name = "GoogLeNet"
model_path = 'models/bvlc_googlenet/'
net_fn   = 'models/bvlc_googlenet/deploy_googlenet_updated.prototxt'
param_fn = model_path + 'bvlc_googlenet.caffemodel'
mean = np.float32([104.0, 117.0, 123.0])

net = caffe.Classifier(net_fn, param_fn,
                       mean = mean, # ImageNet mean, training set dependent
                       channel_swap = (2,1,0)) # the reference model has channels in BGR order instead of RGB

【讨论】：