Calculate the output resolution of a layer

Calculate the output resolution of a layer or pooling operation in a Convolutional Neural Network.

Usage

res_calculate(
  type = c("layer", "pooling"),
  in_res,
  kernel_size,
  stride,
  padding
)

Arguments

type

string. Accepted values are "layer" and "pooling".

in_res

integer. It represents the resolution of the input layer.

kernel_size

integer. It refers to the size of the kernel used in the convolution or pooling operation.

stride

integer. It is the stride length for the convolution or pooling operation. Only used when type is "layer".

padding

integer. It is the amount padding added to the input layer. Only used when type is "layer"

Value

The function returns integer which is the output resolution.

Details

• When type is "layer" the output resolution is calculated as ((in_res - kernel_size + (2 * padding)) / stride) + 1.

• When type is "pooling", the output resolution is calculated as the floor division of in_res by kernel_size.

Examples

if (FALSE) {

# Calculating output resolution for a convolution layer
res_calculate(type = "layer", in_res = 12, kernel_size = 2, stride = 2, padding = 0)

# Calculating output resolution for a pooling layer
res_calculate(type = "pooling", in_res = 12, kernel_size = 2)
}