Table 5.

Main added features in the CNNs.

Network	What’s Novel?
AlexNet	- Apply Rectified Linear Units (ReLU) to add non-linearity
VGG-16	- Deep network, approximately twice as deep as AlexNet
GoogLeNet	- Using dense modules as opposed to stacking convolutional layers
ResNets	- Using batch normalization and skip connections
Inception-ResNet-V2	- Using residual inception blocks instead of inception modules - Adding the inception module (Inception-A) after the stem module - Using more inception modules
DarkNet-19	- Combine Darknet extraction, Network In Network, Inception and Batch Normalization in a single model