conv3d_op

→TensorImplPtr

int conv3d_op(const int & x, const int & W, const int & b, int stride_d, int stride_h, int stride_w, int pad_d, int pad_h, int pad_w, int dilation_d, int dilation_h, int dilation_w, int groups)

source edit

Python wrappers

pylucid.nn.functional.conv3dfunction

Three-dimensional cross-correlation over a batch of volumes.

Performs the natural extension of 2-D convolution to the additional depth axis D, summing the kernel response across $(k_d, k_h, k_w)$ .

Parameters

xTensorImplPtr

Input of shape (B, C_in, D, H, W).

WTensorImplPtr

Filters of shape (C_out, C_in / groups, KD, KH, KW).

bTensorImplPtr

Bias of shape (C_out,) or empty.

stride_d, stride_h, stride_wint= None

Per-axis stride. Default: 1.

pad_d, pad_h, pad_wint= None

Per-axis zero-padding. Default: 0.

dilation_d, dilation_h, dilation_wint= None

Per-axis kernel-element spacing. Default: 1.

groupsint= None

Channel grouping. Default: 1.

Returns

TensorImplPtr

Output of shape (B, C_out, D_out, H_out, W_out) where each output extent follows $X_\text{out} = \lfloor (X + 2 p_x - d_x(K_X - 1) - 1)/s_x + 1 \rfloor$ for $X \in \{D, H, W\}$ .