Data augmentation does not improve network accuracy

[Ali]

Hi,

At the moment, I am using Alexnet to do a classification task.

The size of each input sample is 480*680.

Using a normal network, fed by cropped inputs of size 256*256 (generated in preprocessing steps) with the batch size of 8, gives me the accuracy rate of 92%.

But, when I  try to generate 5 crops of each (480*680) sample (corners plus a center crop) using the following crop layers:

# this is the reference blob of the cropping process which determines cropping size

layer {

  name: "reference-blob"

  type: "Input"

  top: "reference"

  input_param { shape: { dim: 8 dim: 3 dim: 227 dim: 227 } }

}

# upper-left crop

layer{

name: "crop-1"

type: "Crop"

bottom: "data"

bottom: "reference"

top: "crop-1"

crop_param {

axis: 2

offset: 1

offset: 1

}

}

# upper-right crop

layer{

name: "crop-2"

type: "Crop"

bottom: "data"

bottom: "reference"

top: "crop-2"

crop_param {

axis: 2

offset: 1

offset: 412

}

}

# lower-left crop

layer{

name: "crop-3"

type: "Crop"

bottom: "data"

bottom: "reference"

top: "crop-3"

crop_param {

axis: 2

offset: 252

offset: 1

}

}

# lower-right crop

layer{

name: "crop-4"

type: "Crop"

bottom: "data"

bottom: "reference"

top: "crop-4"

crop_param {

axis: 2

offset: 252

offset: 412

}

}

# center crop

layer{

name: "crop-5"

type: "Crop"

bottom: "data"

bottom: "reference"

top: "crop-5"

crop_param {

axis: 2

offset: 127

offset: 207

}

}

# concat all the crop results to feed the next layer

layer{

name: "crop_concat"

type: "Concat"

bottom: "crop-1"

bottom: "crop-2"

bottom: "crop-3"

bottom: "crop-4"

bottom: "crop-5"

top: "all_crops"

concat_param {

    axis: 0

}

}

# generating enough labels for all the crop results

layer{

name: "label_concat"

type: "Concat"

bottom: "label"

bottom: "label"

bottom: "label"

bottom: "label"

bottom: "label"

top: "all-labels"

concat_param {

    axis: 0

}

}

leads to accuracy rate of 90.6% which is strange.

Do you see any mistake in that?