Caffe: Data Layer unknown layer type

[rshanor]

I am trying to follow the example here (https://github.com/BVLC/caffe/blob/master/examples/00-classification.ipynb) to classify one image on a network I have trained. I am currently able to classify batch images using a DenseImageData layer. However I am now trying to classify single images in real time from a camera. Anyway, I now set up my data layer as shown below, using this at reference (https://github.com/BVLC/caffe/blob/master/models/bvlc_reference_caffenet/deploy.prototxt). When I try to call caffe.Net in python, I get the following error: * Error parsing text-format caffe.NetParameter: 7:15: Message type "caffe.LayerParameter" has no field named "input_param"_. When I comment out the input_param layer, I get the following error: _Unknown layer type: Input. *

Thanks for the help!!!

'''layer {
name: "data"
type: "Input"
top: "data"
#top: "label"
input_param { shape: { dim: 1 dim: 3 dim: 195 dim: 185 } }
}'''

[lukeyeager]

How recent is your build? Input layers were added 20 days ago with https://github.com/BVLC/caffe/pull/3211.

[rshanor]

Thanks @lukeyeager that is totally my issue. How would I go about adding this layer type into my caffe build? I am using a custom version of caffe found here: mi.eng.cam.ac.uk/projects/segnet/tutorial.html. Or is there another way to go about classifying a single image in python?

[ShaggO]

I guess you could just use the old definition. Instead of adding a layer, you specify the following directly in the root og the prototxt:
input: "data"
input_dim: 1
input_dim: 3
input_dim: 195
input_dim: 185

Alternatively you can find Alex' github repo here it seems: https://github.com/alexgkendall/caffe-segnet. If you fork the repo, you should be able to add the Caffe repo as remote and fetch the commits. You will have to figure out the details yourself though ;)

[rshanor]

Yup good call. Got it working last night after digging through old commits of the example files. I might try to merge as you suggested, but this will work for now. Thanks @lukeyeager @ShaggO !

[mrgloom]

Same issue, here is solution:

layer {
  name: "data"
  type: "Input"
  top: "data"
  input_param { shape: { dim: 1 dim: 3 dim: 227 dim: 227 } }
}

to

input: "data"
input_shape {
  dim: 1
  dim: 3
  dim: 227
  dim: 227
}

This is very annoying situation with outdated forks and .prototxt format changing in caffe master.

[lukeyeager]

The examples were updated to use the latest features. If you're using an old version of the code, use the examples checked into the repo at that time (e.g. https://github.com/BVLC/caffe/tree/rc3/examples).

The code is backwards-compatible with old examples, but the examples are not backwards-compatible with old code. This is as it should be.

[aligoglos]

i have same problem and try this format :
input: "data"

input_shape {
  dim: 1
  dim: 3
  dim: 640
  dim: 480
}

but don't work!

WARNING: Logging before InitGoogleLogging() is written to STDERR
I0323 16:43:39.494524  6784 common.cpp:36] System entropy source not available, using fallback algorithm to generate seed instead.
I0323 16:43:43.032114  6784 upgrade_proto.cpp:67] Attempting to upgrade input file specified using deprecated input fields: C:/Users/127051/Documents/Visual Studio 2015/Projects/C++/Caffe/CaffeTest/x64/Release/net_struct.prototxt
I0323 16:43:43.033090  6784 upgrade_proto.cpp:70] Successfully upgraded file specified using deprecated input fields.
W0323 16:43:43.033602  6784 upgrade_proto.cpp:72] Note that future Caffe releases will only support input layers and not input fields.
I0323 16:43:43.034595  6784 net.cpp:53] Initializing net from parameters:
state {
  phase: TEST
  level: 0
}
layer {
  name: "input"
  type: "Input"
  top: "data"
  input_param {
    shape {
      dim: 1
      dim: 3
      dim: 640
      dim: 480
    }
  }
}
...


I0323 16:43:44.057319  6784 layer_factory.cpp:58] Creating layer input
F0323 16:43:44.068845  6784 layer_factory.cpp:62] Check failed: registry.count(type) == 1 (0 vs. 1) Unknown layer type: Input (known types: Convolution, Eltwise, LRN, Pooling, Power, Python, ReLU, Sigmoid, Softmax, Split, TanH)
*** Check failure stack trace: ***

[willyd]

Something is wrong with your build. Are you using stock caffe.exe or your own executable?

[willyd]

If using your own executable please add /DCMAKE_WINDOWS_BUILD to your definitions.

[willyd]

Please continue this dicussion on the mailing list.

From https://github.com/BVLC/caffe/blob/master/CONTRIBUTING.md:

Please do not post usage, installation, or modeling questions, or other requests for help to Issues.
Use the caffe-users list instead. This helps developers maintain a clear, uncluttered, and efficient view of the state of Caffe.

[vishalkumarchaudhary]

If unknown layer is imagedata then uncomment this flag OPENCV :=3 in Makefile.config
then compile again ... by

     sudo make clean | sudo  make all