how to handle redundant (not used) blobs in the network?

[springfl...@gmail.com]

Hello,

My input images have 3 channels RGB however I want to process only  two channels R and G. I do not need B.

I use Slice layer to separate the data blob into three blobs: red, green, blue 

layer {
  name: "sliceIm"
  type: "Slice"
  bottom: "data"
  top: "red"
  top: "green"
  top: "blue"
  slice_param {
    slice_point: 1
    slice_point: 2
    axis: 1
  }
}

Only red and green blobs are input to the next layers.

When I trained the network it  only returned a series of:  Test net output #(iteration number): blue = -1.58751 and finally run out of memory.

Can you please advice how I can handle this situation: How to make the network not use the redundant blob "blue"?

Do I have to come back and reprocess the input data so that it has only 2 channels before creating LMDB file? Is there any faster way to handle this situation?

Thank you/

[Evan Shelhamer]

The `Silence` layer type takes bottoms without making any tops to ignore unattached tops like the `blue` top in your example.

Evan Shelhamer

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users...@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/cd0b6d7b-3739-4347-9451-3d4a70e138a4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.