Advanced Deep Learning with Python

Inception v4 and Inception-ResNet

In the latest revision of Inception networks, the authors introduced three new streamlined Inception blocks that build upon the idea of the previous versions (Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning, https://arxiv.org/abs/1602.07261). They introduced 7×7 asymmetric factorized convolutions, and average pooling instead of max pooling. More importantly, they created a residual/Inception hybrid network known as Inception-ResNet, where the Inception blocks also include residual connections. We can see the schematic of one such block in the following diagram:

An Inception ...