Conv2D¶

class Conv2D(out_channels: int, kernel_size: int = 3, stride: int = 1, padding: bool = True)¶

Applies a 2D convolution over and input.

Example:

conv = Conv2D(32, kernel_size=5)
net = NeuralNet([
        ...
        conv,
        ...
    ])

__init__(out_channels: int, kernel_size: int = 3, stride: int = 1, padding: bool = True)¶

Initializes the object.

Parameters

out_channels (int) – number of output channels
kernel_size (int, optional) – size of the window, defaults to 3
stride (int, optional) – determens how much the filter moves, defaults to 1
padding (bool, optional) – True if you want to add padding to the input image, defualts to True

Methods

`__init__`(out_channels[, kernel_size, …])	Initializes the object.
`backward`(grad)	Used to calculate the gradients of weights and biases.
`build`(in_channels)	Builds the layer.
`forward`(inputs[, training])	Called each time the data passes throughout the nework.
`get_params`()	Returns params of layer.
`remove_zero_padding`(inp)	Add zero padding to the input Tensor.
`step`(inp, w, b)	Perform single step in convolution.
`zero_padding`(inp)	Add zero padding to the input Tensor.

backward(grad: numpy.ndarray) → numpy.ndarray¶

Used to calculate the gradients of weights and biases.

Parameters: grad (Tensor) – gradient from previous layer or loss function.
Returns: Gradient
Return type: Tensor

build(in_channels: int)¶

Builds the layer.

forward(inputs: numpy.ndarray, training: bool = False) → numpy.ndarray¶

Called each time the data passes throughout the nework.

Parameters

inputs (Tensor) – output from the previous layer
training (bool, optional) – set to true during training, and is false when network predicts

Returns

processed input data

Return type

Tensor

get_params() → Dict¶

Returns params of layer.

remove_zero_padding(inp: numpy.ndarray) → numpy.ndarray¶

Add zero padding to the input Tensor.

step(inp: numpy.ndarray, w: numpy.ndarray, b: numpy.ndarray) → numpy.ndarray¶

Perform single step in convolution.

Parameters

zero_padding(inp: numpy.ndarray) → numpy.ndarray¶

Add zero padding to the input Tensor.

BlueBird