Deep Convolutional Neural Networks for Chest Diseases Detection

Table 3

CNN training parameters.

Layers	Description	Values
Input layer	Input image	32 × 32 × 1 images with “zerocenter” normalization
Hidden layer 1	Conv1 + BN + ReLu	16 feature maps of size 10 × 10
	Pool1	2 × 2 kernel size with stride of 2
Hidden layer 2	Conv2 + BN + ReLu	32 feature maps of size 10 × 10
	Pool2	2 × 2 kernel size with stride of 2
Hidden layer 3	Conv3 + BN + ReLu	64 feature maps of size 10 × 10
Classification layer	FC	2 fully connected layers
	Softmax	2 units