Utility Layers

class ConcatLayer

Concating multiple blobs into one along the specified dimension. Except in the concatenation dimension, the shapes of the blobs being concatenated should be the same.


Default 3 (channel). The dimension to concat. Should be an iteger within the range 1 to 4.


Names of the blobs to be concatenated.


Name of the concatenated output blob.

class HDF5OutputLayer

Take some blobs in the network and write the blob contents to a HDF5 file. Note the target HDF5 file will be overwritten when the network is first constructed, but later iterations will append data for each mini-batch. This is useful for storing the final predictions or the intermediate representations (feature extraction) of a network.


The path to the target HDF5 file.


Default false. When the layer tries to create the target HDF5 file, if this attribute is enabled, it will overwrite any existing file (with a warning printed). Otherwise, it will raise an exception and refuse to overwrite the existing file.


A list of names of the blobs in the network to store.


Default []. Should either be empty or a list of Symbol of the same length as bottoms. Each blob will be stored as an HDF5 dataset in the target HDF5 file. If this attribute is given, the corresponding symbol in this list is used as the dataset name instead of the original blob’s name.

class ReshapeLayer

Reshape a blob. Can be useful if, for example, you want to make the flat output from an InnerProductLayer meaningful by assigning each dimension spatial information.

Internally there is no data copying going on. The total number of elements in the blob tensor after reshaping should be the same as the original blob tensor.


Default 1. The new width after reshaping.


Default 1. The new height after reshaping.


Default 1. The new channels after reshaping.


Blob names for output and input. This layer could take multiple input blobs and produce the corresponding number of output blobs. The shapes of the input blobs do not need to be the same. But the feature dimensions (product of the first 3 dimensions) should be the same.

class SplitLayer

Split layer produces identical copies [1] of the input. The number of copies is determined by the length of the tops property. During back propagation, derivatives from all the output copies are added together and propagated down.

This layer is typically used as a helper to implement some more complicated layers.


Input blob names, only one input blob is allowed.


Output blob names, should be more than one output blobs.

[1]All the data is shared, so there is no actually data copying.