Deeper architecture cnn

#1

Hi,
How can we make any cnn architecture deeper? like eg.- we want to make vgg architecture deeper?
Thanks

#2

Hi,
By adding more convolutional layers or pooling layers and fully connected layers you can make it deeper like VGG16 has 16 (11 cnn+5 pooling) layers and VGG19 has 19 layers.

Add the layers, till you get the required accuracy, but there is a saturations level after which adding more layers , will have no affect on the accuracy. There you need to stop.

RESNET 152 by Kaming He in 2015 in this and CUIMAGE also called Sqeeze and Excitations Networks"(modified RESNET) already go the top-5 error rate 2.251. later this the more CNN addition has not helped as the accuracy and performance got saturated,

So, you can use this CNN for your classifications task.

All the best!