Extensible Embedded Processor for Convolutional Neural Networks

Table 3

Configuration 3: CNN model summary.

Convert all layers
Layer (type)	Kernel size	Output shape	Parameters
conv2d_1	3 × 3	(126, 126, 18)	180
conv2d_2	3 × 3	(124, 124, 18)	2934
conv2d_3	3 × 3	(122, 122, 18)	2934
conv2d_4	3 × 3	(120, 120, 18)	2934
max_pooling2d_1	2 × 2	(60, 60, 18)	0
max_pooling2d_2	2 × 2	(30, 30, 18)	0
conv2d_5	3 × 3	(28, 28, 36)	5868
conv2d_6	3 × 3	(26, 26, 36)	11700
max_pooling2d_3	2 × 2	(13, 13, 36)	0
max_pooling2d_4	2 × 2	(6, 6, 36)	0
conv2d_7	3 × 3	(4, 4, 120)	39000
flatten_1 (flatten)		(1, 1920)	0
dense_1 (dense)		(1, 120)	230520
dense_2 (dense)		(1, 10)	1210
		Total parameters:	297280