CReorgLayer Class

June 14, 2020 ยท View on GitHub

This class implements a layer that transforms a set of two-dimensional multi-channel images into a set of images of smaller size but with more channels. This operation is used in YOLO architecture.

Settings

Resize factor

void SetStride( int stride );

Sets the value by which the image size will be divided in the final result. The image size along either dimension should be a multiple of this value. The value should be greater than 1.

Trainable parameters

The layer has no trainable parameters.

Inputs

The single input accepts a blob with the images, of the dimensions:

  • BatchLength * BatchWidth * ListSize is equal to the number of images
  • Height is the image height; should be a multiple of GetStride()
  • Width is the image width; should be a multiple of GetStride()
  • Depth is equal to 1
  • Channels is the number of channels in the image format

Outputs

The single output contains a blob with the resulting images, of the dimensions:

  • BatchLength is equal to the input BatchLength
  • BatchWidth is equal to the input BatchWidth
  • ListSize is equal to the input ListSize
  • Height is equal to the input Height / GetStride()
  • Width is equal to the input Width / GetStride()
  • Depth is equal to 1
  • Channels is equal to the input Channels * GetStride() * GetStride()

Each image in the set is split in the same way as the following sample.

Assume we have a 2-channel image 4 by 6, and GetStride() is 2. The image pixel values are:

// First channel contents
1,  2,  3,  4,  5,  6,
7,  8,  9,  10, 11, 12,
13, 14, 15, 16, 17, 18,
19, 20, 21, 22, 23, 24,

// Second channel contents
25, 26, 27, 28, 29, 30,
31, 32, 33, 34, 35, 36,
37, 38, 39, 40, 41, 42,
43, 44, 45, 46, 47, 48

This image will be transformed into a 8-channel image 2 by 3, with the contents:

// First channel contents
1,  3,  5,
7,  9,  11,

// Second channel contents
25, 27, 29,
31, 33, 35,

// Third channel contents
13, 15, 17,
19, 21, 23,

// Fourth channel contents
37, 39, 41,
43, 45, 47

// Fifth channel contents
2,  4,  6,
8,  10, 12,

// Sixth channel contents
26, 28, 30,
32, 34, 36,

// Seventh channel contents
14, 16, 18,
20, 22, 24,

// Eighth channel contents
38, 40, 42,
44, 46, 48