Table 1.

Feature embedding CNN architecture.

Layer	No. Filters
$2\times$ Conv2D	8
MaxPooling2D	−
Batch Normalisation	−
$2\times$ Conv2D	16
MaxPooling2D	−
Batch Normalisation	−
$3\times$ Conv2D	32
MaxPooling2D	−
Batch Normalisation	−
$3\times$ Conv2D	64
MaxPooling2D	−
Batch Normalisation	−
Flatten	−

The size of the kernels is identical for all convolutional layers and is set to $3\times 3$, with the convolutional stride set to $1\times 1$. Max-pooling is performed after each block of convolutional layers over a $2\times 2$ window, with a $2\times 2$ stride.