Fig. 17: Separable-convolution scheme: different convolution kernels are applied along the lateral x, y axes vs. the longitudinal z-axis. | Light: Science & Applications

Fig. 17: Separable-convolution scheme: different convolution kernels are applied along the lateral x, y axes vs. the longitudinal z-axis.

From: Dynamical machine learning volumetric reconstruction of objects’ interiors from limited angular views

Fig. 17: Separable-convolution scheme: different convolution kernels are applied along the lateral x, y axes vs. the longitudinal z-axis.

In our present implementation, the kernels’ respective dimensions are 3 × 3 × 1 (or 1 × 1 × 1) and 1 × 1 × 4. The lateral and longitudinal convolutions are computed separately and the results are then added element-wise. The separable convolution scheme is used in both the gated recurrent unit and the encoder/decoder (for more details, see Convolutional encoder and decoder in Materials and methods)

Back to article page