Extensible Embedded Processor for Convolutional Neural Networks

Table 1

Configuration 1: CNN model summary.

Layer (type)	Original MSTAR model	Output shape	Parameters
	Kernel size
conv2d_1	9 × 9	(120, 120, 18)	1476
max_pooling2d_1	6 × 6	(20, 20, 18)	0
conv2d_2	5 × 5	(16, 16, 36)	16236
max_pooling2d_2	4 × 4	(4, 4, 36)	0
conv2d_3	4 × 4	(1, 1, 120)	69240
flatten_1 (flatten)		(1, 120)	0
dense_1 (dense)		(1, 120)	14520
dense_2 (dense)		(1, 10)	1210
		Total parameters:	102682