Error encountered: Unknown layer type encountered: 4 for VGG-16 laycer CNN

[StephenBydawell]

Using the command: ./convert.py VGG_ILSVRC_16_layers_deploy.prototxt --code-output-path=mynet.py

.prototxt obtained from: https://gist.github.com/ksimonyan/211839e770f7b538e2d8

[SepehrV]

The problem is with the version of Caffe that your prototxt file is written for.
In more recent versions of Caffe, some syntax for prototxt files has changed. E.g. layers changed to layer or now the type of layers is a string.
So to solve the issue you can simply modify the prototxt file so that it would be compatible with latest release of Caffe.

To do so, apply these changes:

layers -> layer
CONVOLUTION -> "Convolution"
POOLING -> "Pooling"
RELU -> "ReLU"
INNER_PRODUCT -> "InnerProduct"
SOFTMAX -> "Softmax"
DROPOUT -> "Dropout"
LRN -> "LRN"

[StephenBydawell]

Thanks for your response. I implemented the changes and now have this error:

xt_format.py", line 764, in ConsumeIdentifier
raise self._ParseError('Expected identifier.')
google.protobuf.text_format.ParseError: 1:1 : Expected identifier.

I changed the attached .prototxt to .txt
VGG_ILSVRC_16_layers_deploy.txt

Your assistance is much appreciated.

[ethereon]

I'd recommend using the upgrade_net_proto_text and upgrade_net_proto_binary tools bundled with Caffe for updating the .prototxt and .caffemodel files respectively, as described in the readme.

[StephenBydawell]

Managed to get a new .prototxt. Thanks for the help.

[Yaffa1607]

Hi , I do not have caffe , I was wondering if someone can upgrade a prototxt file and share?
I'm trying to convert this caffe model to tensprflow. http://vision.soic.indiana.edu/egohands_files/hand_classifier.prototxt

[Serhiy-Shekhovtsov]

@Yaffa1607 have you ever got the hand_classifier.prototxt converted?

[geoffzhang]

Hi, @SepehrV , Do you have the problem of "Error encountered: Unknown layer type encountered: Normalize", can you give me some advice, please?

[sumaira-hussain]

Hi
I am using same command to convert my caffe model into tensorflow. I have updated my both prototxt and caffemodel files. but still I am getting this error.

can anybody tell what is wrong?