Human oocytes image classification method based on deep neural networks

Table 4 Network structure

Name	Type	Activations	Parameters
inputLayers	Input	590 × 590 × 2
conv1	Convolution	294 × 294 × 32	Conv 2D 3 × 3 stride [2 2] padding [0 0 0 ]
relu_conv1	ReLu	294 × 294 × 32
conv2	Convolution	146 × 146 × 16	Conv 2D 1 × 1 stride [1 1] padding [0 0 0]
relu_conv2	ReLu	146 × 146 × 16
pool1_conv2	Max pooling	72 × 72 × 16	3 × 3 stride [2 2] padding [0 0 0]
conv4	Convolution	72 × 72 × 8	Conv 2D 1 × 1 stride [1 1] padding [0 0 0]
relu_conv4	ReLu	72 × 72 × 8
res_conv5_1	Convolution	72 × 72 × 28	Conv 2D 3 × 3 stride [1 1] padding [0 0 0]
relu_conv5_1	ReLu	72 × 72 × 28
res_conv5_2	Convolution	72 × 72 × 28	Conv 2D 1 × 1 stride [1 1] padding [0 0 0]
relu_conv5_2	ReLu	72 × 72 × 28
depth_concat1	Depth concatenation	146 × 146 × 56
pool_depth_concat1	Max pooling	36 × 36 × 56	3 × 3 stride [2 2] padding [0 1 0 1]
conv6	Convolution	36 × 36 × 4	Conv 2D 1 × 1 stride [1 1] padding [0 0 0]
relu_conv6	ReLu	36 × 36 × 4
conv7	Convolution	36 × 36 × 4	1000 1 × 1 × 48 stride [1 1] padding [0 0 0]
relu_conv7	ReLu	36 × 36 × 4
global_pool	2D global avg pooling	1 × 1 × 4
fullyConnected1	Fully connected	3 outputs
softMax_output	Softmax	3

ISSN: 1475-925X