Research Article

Extensible Embedded Processor for Convolutional Neural Networks

Table 2

Configuration 2: CNN model summary.

Keep 6 × 6 max pooling layer
Layer (type)Kernel sizeOutput shapeParameters

conv2d_13 × 3(126, 126, 18)180
conv2d_23 × 3(124, 124, 18)2934
conv2d_33 × 3(122, 122, 18)2934
conv2d_43 × 3(120, 120, 18)2934
max_pooling2d_16 × 6(20, 20, 18)0
conv2d_53 × 3(18, 18, 36)5868
conv2d_63 × 3(16, 16, 36)11700
max_pooling2d_22 × 2(8, 8, 36)0
max_pooling2d_22 × 2(4, 4, 36)0
conv2d_73 × 3(2, 2, 120)39000
flatten_1 (flatten)(1, 480)0
dense_1 (dense)(1, 120)57720
dense_2 (dense)(1, 10)1210

Total parameters:124480