Table 1.

Data dimensions in CORONA-Net (the backbone/encoder network).

Layers	Kernel Size	Stride	Padding	Output Channel
$Input$	Tensor (50, 3, 224, 224 )	-	-	-
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 64
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 64
$MaxPooling$	K = 2 × 2	S = 2	P = 0	-
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 128
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 128
$MaxPooling$	K = 2 × 2	S = 2	P = 0	-
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 256
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 256
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 256
$MaxPooling$	K = 2 × 2	S = 2	P = 0	-
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 512
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 512
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 512
$MaxPooling$	K = 2 × 2	S = 2	P = 0	-
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 512
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 512
$Convolution$ $(BatchNorm+Relu)$	K = 3 × 3	S = 1	P = 1	O = 512
$Convolution$	K = 1 × 1	S = 1	P = 0	O = 100